IJACSIT Volume 03 - ELVEDIT.COM

Contents

Editorial Board I-IV

Holistic Electronic Government Services Integration Model: from Theory to Practice Tadas Limba, Gintarė Gulevičiūtė 1-31

Detecting Suspicion Information on the Web Using Crime Data Mining Techniques Javad Hosseinkhani, Mohammad Koochakzaei, Solmaz Keikhaee, Yahya Hamedi Amin 32-41

Develop a New Method for People Identification Using Hand Appearance Mahdi Nasrollah Barati, Seyed Yaser Bozorgi Rad 42-49

Model and Solve the Bi-Criteria Multi Source Flexible Multistage Logistics Network Seyed Yaser Bozorgi Rad, Mohammad Ishak Desa, Sara Delfan Azari 50-69

Impact of Strategic Management Element in Enhancing Firm’s Sustainable Competitive Advantage. An Empirical Study of Nigeria’s Manufacturing Sector Yahaya Sani, Abdel-Hafiez Ali Hassaballah 70-82

A Co-modal Transport Information System in a Distributed Environment Zhanjun Wang, Khaled Mesghouni, Slim Hammadi 83-99

Online Brand Experience Creation Process Model: Theoretical Insights Tadas Limba, Mindaugas Kiskis, Virginija Jurkute 100-118

Color Image Segmentation Using a Modified Fuzzy C-means Method and Data Fusion Techniques Rafika Harrabi, Ezzedine Ben Braiek 119-134

Model of Brand Building and Enhancement by Electronic Marketing Tools: Practical Implication Tadas Limba, Gintarė Gulevičiūtė, Virginija Jurkutė 135-155 A Constraint Programming Approach for Scheduling in a Multi-Project Environment Marcin Relich 156-171 Anti-Crisis Management Tools for Capitalist Economy Alexander A. Antonov 172-190 E-Business Qualitative Criteria Application Model: Perspectives of Practical Implementation Tadas Limba, Gintarė Gulevičiūtė 191-213 A UML Profile for use cases Multi-interpretation Mira Abboud, Hala Naja, Mohamad Dbouk, Bilal El Haj Ali 214-226 A Grid-enabled Application for the Simulation of Plant Tissue Culture Experiments Florence I. Akaneme, Collins N. Udanor, Jane Nwachukwu, Chibuike Ugwuoke, Carl .E.A Okezie, Benjamin Ogwo 227-242 Clustering Evolutionary Computation for Solving Travelling Salesman Problems Tanasanee Phienthrakul 243-262 An Agent Driven M-learning Application Collins N. Udanor, O.U. Oparaku 263-272 Evolution of Utilizing Multiple Similarity Criteria in Web Service Discovery Hassan Azizi Darounkolaei, Seyed Yaser Bozorgi Rad 273-281 Multi-Aspect Tasks in Software Education: A Case of a Recursive Parser Evgeny Pyshkin 282-305 Structured Stream Data Mining Using SVM Method as Basic Classifier Hadi Barani Baravati, Javad Hosseinkhani, Solmaz Keikhaee, Abbas Shahsavari 306-316

Models for Integrating Social Networking in Higher Education Andreas Veglis 317-326 Wireless Sensor System According to the Concept of IoT -Internet of Things- Juan Felipe Corso Arias, Yeison Julian Camargo Barajas, Juan Leonardo Ramirez Lopez 327-343 Analysis of Multiple String Pattern Matching Algorithms Akinul Islam Jony 344-353 Nonlinearity Compensation for High Power Amplifiers Based on Look-Up Table Method for OFDM Transmitters Maryam Sajedin, Ayaz Ghorbani, Hamid Reza Amin Davar 354-367 E-Portfolio Assessment for Learning: Ten Years Later – an Experience from an Outcome-Based University Abdallah Tubaishat 368-378 How Programmer Plans Training? Jakub Novotný, Martina Winklerová 379-389

I

Editorial Board

Dr. Tanveer A Zia

• Associate Head of School of Computing and Mathematics

Charles Sturt University, Australia

Prof. Dr. Sun-Yuan Hsieh

• Head of Department of Computer Science and Information Engineering

National Cheng Kung University, Taiwan

Prof. Dr. Loet Leydesdorff

• Professor of Amsterdam School of Communication Research (ASCoR)

University of Amsterdam, Netherlands

Dr. Smain Femmam

• Associate professor of Strasbourg University of Haute Alsace

Associate member in the Laboratory of Signals and Safety Systems of Polytechnic School of Engineering,

France

Prof. Dr. Milena M. Head

• Professor Information Systems,

Acting Director MBA Programs

DeGroote School of Business

McMaster University, Canada

Prof. Dr. DV Ashoka

• Head and Professor of Department of Information Science and Engineering

Visveswaraya Technological University, India

Prof. Dr. Li-Der Chou

• Department of Computer Science and Information Engineering

National Central University, Taiwan

II

Dr. Khaled Hassanein

• Professor of Information Systems

Chair, Information Systems Area

Director, McMaster eBusiness Research Centre (MeRC)

DeGroote School of Business

McMaster University, Canada

Dr. Ching-Hsien Hsu

• Department of Computer Science and Information Engineering

Chung Hua University, Taiwan

Prof. Valentina Emilia Balas

• Faculty of Engineering

Aurel Vlaicu University of Arad, Romania

Dr. Hamed Taherdoost

• CEO of Ahoora Ltd | Management Consultation Group

Head of R&D Department, AsanWare Sdn Bhd, Malaysia

Dr. Fu-Hau Hsu

• Computer Science and Information Engineering Department


Prof. Dr. P.Raviraj

• Head and Professor of Computer Science and Engineering Department

Kalaignar Karunanidhi Institute of Technology, India

Dr. Yung-Pin Cheng

• Associate Professor of Computer Science and Information Engineering Department


Dr. Ahmed Nabih Zaki Rashed

• Electronics and Electrical Communication Engineering Department, Faculty of Electronic Engineering

Menoufia University, Egypt

Dr. Weiling Ke

• School of Business

Clarkson University, USA

III

Dr. Vusa Sreenivasarao

• Faculty of Electrical and Computer engineering

Bahir Dar University, Ethiopia

Prof. Dr. Anil Kumar

• Department of Mathematics

Greater Noida College of Technology, India

Dr. Ebtesam Najim Abdullah AlShemmary

• Head and Professor of Informatics Center for Research and Development

University of Kufa, Iraq

Dr. Ashu Gupta

• School of Information Technology

Apeejay Institute of Management Technical Campus, India

Dr. Chung-Cheng Chiu

• Department of Electrical and Electronic Engineering

Chung Cheng Institute of Technology

National Defense University, Taiwan

Dr. MV Raghavendra

• Associate Professor of Electronic and Communication Department

Adama Science and Technology University, Ethiopia

Dr. Kalpana Chauhan

• Head of Department of EEE

Srida Group of Institutions, India

Dr. Prasant Singh Yadav

• Dean and Associate Professor of Vedant Institute of Management and Technology, India

Dr. Rajender Bathla

• Department of Computer Science and Engineering

Haryana Institute of Engineering and Technology

Kurukshetra University, India

Dr. Rajesh Timane

• MBA Department

IV

Panjabrao Deshmukh Institute of Management Technology and Research

Management Department of Dhanwate National College

NagpurRashtrasant Tukadoji Maharaj Nagpur University, India

Dr. Cheng-Hsiang Liu

• Head of Academic Department Industrial Management

National Pingtung University of Science and Technology, Taiwan

Prof. Dr.Pan Quan-Ke

• State Key Laboratory of Synthetical Automation for Process Industries,

Northeastern University, China

Prof. Dr. Mohd Nazri Ismail

• National Defence University of Malaysia, Malaysia

Prof. Dr. Haider M. AlSabbagh

• Department of Electrical Engineering

College of Engineering

University of Basra, Iraq

Prof. Dr. Muzhir Shaban Al-Ani

• College of Computer

University of, Iraq

Dr. Ricardo Rodriguez

• Department of Mechatronics

Technological University of Ciudad Juarez, Mexico

Prof. Dr. Vida Davidavičienė

• Head of Department of Business Technologies

Vilnius Gediminas Technical University, Lithuania

Dr. Wanqing Tu

• Associate Professor

Department of Computing Science and Digital Media

The Robert Gordon University, Aberdeen, UK

International Journal of Advanced Computer Science and Information Technology (IJACSIT) Vol. 3, No. 1, 2014, Page: 1-31, ISSN: 2296-1739 © Helvetic Editions LTD, Switzerland www.elvedit.com

Holistic Electronic Government Services Integration Model: from Theory to Practice

Authors

Tadas Limba Institute of Digital Technologies, Faculty of Social Technologies, Mykolas Romeris University

[email protected] Vilnius, LT-0100, Lithuania

Gintarė Gulevičiūtė Institute of Digital Technologies, Faculty of Social Technologies, Mykolas Romeris University


Abstract

The systematic, comparative analysis of the models of electronic government services carried out in the scientific work and the assessment of opportunities of their application in the self-government level makes the topic a novelty. With the help of the method of comparative analysis the models of electronic government services have been assessed and there has been distinguished the total of six. Two of them being the main common models of electronic government services have the features that enable the development of new models of electronic government services that are more targeted at changes taking place in public needs and inside organizational processes signifying the originality. The aim of this work is to develop a Holistic Electronic Government Services Integration Model which could ensure the efficient integration of electronic government services in the local self-government level. The scientific work analyzes the improvement opportunities of the models of electronic government services and their application alternatives in Lithuanian municipalities. In order to evaluate implementation of “Holistic Electronic Government Services Integration Model”, four empirical studies have been conducted, which show the possibility of this model application. The newly developed model of electronic government services that has been designed basing on the principle of integrating online expert consultation is primarily targeted at improvement of inside processes’ changes of an organization. Practicing the application of that model in the local self-government level starting with improvement of inside processes of an organization should help adapt more accurately and efficiently to the changing needs of the society while providing electronic government services, thus establishing a higher public value. The practical novelty of work is reflected not only through the integration opportunities’ assessment of the principle of online expert consultation services into the theoretical models of electronic government services that have already been developed by the scientists, but also on the basis of this principle there has been created a “Holistic Electronic Government Services Integration Model” in accordance with “E-Diamond” model basis and its practical application realization with the design of “The project of implementing the principle of online expert consultation on the model of electronic government services” for the future investigations.

International Journal of Advanced Computer Science and Information Technology Volume 3, Issue 1, 2014, ISSN: 2296-1739

Copyright © 2014 Helvetic Editions LTD - All Rights Reserved

www.elvedit.com 2

Key Words

E-government, electronic government services, Stage model of electronic government services, “E-Diamond” model of electronic government services, Holistic Electronic Government Services Integration Model, local authorities, public administration.

I. INTRODUCTION Electronic government is being implemented worldwide and in all levels of governance. Local

authorities play a particularly important role, since they can identify the changing customers’ needs best of all [1]. Although the governance functions of those institutions in different countries slightly differ, their common goal remains the same, that is not only to make the governance itself more efficient, but also to make it more accessible for the public [17, 18]. Municipalities, that are able to provide public service for the consumers in more effective and modern ways can also meet other public needs, in this way implementing directly one of the principles of European local self-government charter, which is the one of ensuring a tighter link between local self-government and the public. Basing on that it can be claimed, that the role of electronic government is of quite an important manner while making an impact on a suitable choice of different models for the implementation of the above mentioned and other principles of local authorities.

Scientific issue. Issues and their solutions concerning the efficient electronic government services provision occur worldwide. In some countries, Lithuania is one of them, the models that are identified only hardly coordinate with the models of electronic government services covered in this work or separate fragments of such models being applied in local self-government levels. In order to solve those problems it would be expedient to find new, improved and more effective models of electronic government services that can meet the needs of customers better while providing electronic government services.

Object of the research. The application of the models of electronic government services for public administration.

Purpose – the aim of work is to develop a Holistic Electronic Government Services Integration Model which could ensure the efficient integration of electronic government services in the local self-government level.

There have been set the following objectives for the above mentioned purpose to be achieved: 1. To carry out the comparative analysis of Stage models of electronic government services; 2. To analyze in detail “E-Diamond” electronic government services model; 3. Having analyzed the Stage models and “E-Diamond” model of electronic government

services and having carried out their comparative analysis, to establish the main features of those models for their improvement;

Holistic Electronic Government Services Integration Model: from Theory to Practice Tadas Limba, Gintarė Gulevičiūtė


www.elvedit.com 3

4. To reveal the conceptual-holistic prospects for improving electronic governance services models;

5. To offer a new model of electronic government services for the local authorities; 6. To conduct research of application a new model of electronic government services in

Lithuanian municipalities.

Practical significance. The development alternatives of the suggested Holistic Electronic Government Services Integration Model and its principle of online expert consultation services for municipalities provide the conditions to carry out the experiment thus pointing out the practical value of the scientific work.

Models of electronic government services, that are created and being analyzed by worldwide scientists, aim at seeking solutions and their alternatives for more efficient public services provision. The years of establishment of the models reflecting the recent development trends of models of electronic government services show that the subject matter of the models of electronic government services is relevant and fairly new worldwide as well as in Lithuania.

II. ANALYSIS OF STAGES MODELS OF ELECTRONIC GOVERNANCE SERVICES An easy way to comply with the paper formatting requirements is to use this document as a

template and simply type your text into it. All Stages Models – „ANAO“, “SAFAD”, „Lee & Layne“, Public sector processes completing („PSP completing) and „Hiller & Belanger“, which are already analyzed by G. Goldkuhl and A. Persson, have a common feature, namely the first stage has poor functionality and the last one has low level of integration involving all management levels (local and other authorities, legal and natural persons). Another common feature comes with the level of integration, i.e. the higher it is the higher requirements for technologies [23]. Given the common features there are some clear differences as well. In order to compare the models in a more convenient manner the similarities and differences have been put in a table. The features of the models are presented in the columns and the rows list the features of the stages (see Table 1).

TABLE I: A COMPARATIVE ANALYSIS OF THE STAGES MODELS

„ANAO“ “SAFAD” „Lee & Layne“ „PSP completing“ „Hiller & Belanger“

I Publishing information

Information Catalogue Cultivation Information

II Interaction Interaction Transaction Extention Two-way Communication

III Transaction of secure information

Transaction Vertical integration

Maturity Transaction

IV Sharing information with other agencies

Integration Horizontal integration

Revolution Integration

V (Does‘nt exist) (Does‘nt exist) (Does‘nt exist) (Does‘nt exist) Political participation Source: Limba, T., 2009.



www.elvedit.com 4

The comparative analysis of the models that has been carried out shows the fact that electronic

governance stage models have advantages and disadvantages although it is worth remembering that in both theoretical and practical works, where the issues of electronic governance are being analyzed, we would not possibly find absolutely perfect models that would be able to fit any public administration systems. A common spread of stage models in different worldwide public sector systems shows that this type of models are easily implemented in public administration, although it is necessary to emphasize that the spread of their application does not completely illustrate the usefulness and quality of the models while providing public service for users. Having evaluated the weaknesses and threats of the models the essential issues must be pointed out. Firstly, integration of stages is rather complicated process, e.g. a higher stage cannot be integrated without firstly integrating a lower stage. However, this is mainly the issue of technological solutions. Another more important problem might arise while implementing stage models at the moment, namely the rapid change in public sector organizational processes and increasing aiming at individual and complex needs of users, which are related to multiple social phenomena. In conclusion, it can be stated that having assessed the advantages and disadvantages of the above mentioned models’ application in public sector, there could be explored not only those but also more developed alternatives for implementing electronic governance models in public administration institutions.

III. ANALYSIS OF “E-DIAMOND” ELECTRONIC GOVERNANCE SERVICES MODEL All paragraphs must be indented. All paragraphs must be justified, i.e. both left-justified and

right-justified. Electronic governance service model “E-Diamond” is based on a different attitude to public service in comparison with the analyzed stage model group, although it is not completely distant from stage models. This model has been designed as an opposition to stage models. On the other hand, the authors themselves believe that the development of the “E-Diamond” model is restructuring the stage models into poles [16]. The analysis of the “E-Diamond” model and efforts to find connections with stages can be based on the Swedish “SAFAD” stage model [15].

The analysis of stage models shows the predominance of aiming to collect informative services into one pool around one particular subject, in other words, coordinated, individualized and informative electronic governance services (further on – services). The following opposition for this classification is provided in the E- Diamond model: separation, generalizing and performativeness. This kind of division composes three electronic governance services poles [15]. The poles with subdivisions are illustrated in the picture below (see Figure 1).



www.elvedit.com 5

FIGURE I: “E-DIAMOND” MODEL. SOURCE: GOLDKUHL, G., PERSSON, A., STURGART, 2006, P. 72

The first opposition in the poles is separate and coordinated services. Separate meaning the

ones that require only one institution to provide service, their opposition being coordinated services that correspond stage model integration levels [15] and in a way the partial services layer of the interinstitutional multiple interaction model structure [18]. Coordinated services have two more subdivisions: aligned and fused services.

The second opposition includes general and individualized services. General services are provided for everyone without requiring personal identification in the database, whereas individualized services are provided only having identified a person in the electronic space. The latter are divided into non-secured and secured services. Non-secured services in this model can be related with electronic mail requests in stage models (the second stage of “SAFAD”) where you need the basic personal identification (basic presentation indicating the name, surname, occupation, etc.), but the information provided for the applicant is not personal, thus a secured personal identification is not necessary, which would be extremely important for secured services subdivision under the fourth stage of “SAFAD”.

The last poles of the opposition provide informative and performative services. Informative services are targeted only at the information available for reading, e.g. information provided in the website of municipality, whereas the performative services allow the user pursue interaction with an institution on the integrational level [15,16]. For instance, the following information is available on the website of municipality after submitting application for it: information on the vision, mission, objectives and activity. Further systemized information and valid public services are provided after having clarified the request if it is required or carrying out different interinstitutional procedures on the computerized databases. It is worth mentioning that informative services can be subdivided into pre-arranged knowledge and selected information. Pre-arranged information (or knowledge) is received when a customer is surfing the website of an



www.elvedit.com 6

institution, and selected information is the one that is filtered through the browsing system on the site [16].

The founders of the “E-Diamond” model criticize the negative application of the stages principle [16]. The stage models aim at including all the previous stages to the last stage and they set it as the objective, although it is not always possible and necessary to transmit services from the first level to the highest, e.g. there is no point in seeking to transmit the publicity of municipality documents to the highest level. Here the advantages of the “E-Diamond” model come up. This model does not aim at transmitting everything to the highest level, because there is no such level. There are three incomparable areas that are all of equal importance and necessary, and the complexity of services fluctuates in different poles of the areas [16].

Assessing this model according to the analysis that has been carried out firstly it is necessary to point out that the practice of applying the “E-Diamond” model is not very common. As it was mentioned above, the model has been developed by Swedish scientists, so it has been tried to be applied in Sweden. Having assessed the theoretical analysis of the “E-Diamond” model, its application should be easier implemented in practice from the technologial point of view than the stage models, because in that case the barriers are created due to the pursue of the higher stage. However, from the organizational processes development point of view, the implementation of the “E-Diamond” model might be a little bit more complicated for a few reasons. Firstly, the lack of experience of implementing such a model might create certain obstacles for implementation and proper sorting out of public services to particular poles. As it was analyzed describing the threats for model implementation, it might occur that at the same time one or a few services should be attributed to different opposite poles.

However, it is more likely that the main problem is related to the lack of expertise of public administration institutions civil servants in the area of electronic governance project impplementation field. The “E-Diamond” model structure in itself is slightly more complicated than that of stage models, therefore it could be harder taken in and brought into awareness by the civil servants, responsible for implementing projects of this kind in the public sector.

IV. COMPARATIVE ANALYSIS OF THE STAGES AND “E-DIAMOND” ELECTRONIC GOVERNANCE SERVICES MODELS: AN IMPACT FOR THEIR IMPROVEMENT

All electronic governance models that have been analysed in this paper have their own distinguishing features in comparison with other models. The stages and „E-Diamond“ models are considered universal. Stage Models, e.g., „SAFAD“, „ANAO“, are considered to be founding models, others basing on the origin or their paradigm are complementing the founding ones or emphasize other features of the models – “E-Diamond” Model. What is also necessary to point out is that electronic governance Stage Models that have been developed and applied for a longer period of time, have been useful for a while, although further application prospects cast certain scientific doubts. On the other hand, what is also arguable is the usefulness of later developed theoretical electronic governance models (the case of „E-Diamond Model).



www.elvedit.com 7

In order to assess the usefulness and more suitable application of electronic governance models in public sector in a more accurate and efficient manner, one of the proposed alternatives is to define the common features of electronic governance Stages Model and “E-Diamond” Model, which could be the following:

Possible levels of implementation The main features of different level, or in case they are not available, different stages

or steps; The level of targeting at the client; The level of targeting at organizational inner processes; Feedback (self-assessment opportunity); Technological background for the implementation of the selected model [23].

The comparative analysis according to the features mentioned above will also help distinguish the advantages and disadvantages of the model (see Table 2).

TABLE II: A COMPARATIVE ANALYSIS OF COMMON FEATURES OF THE STAGES AND “E-DIAMOND” ELECTRONIC GOVERNANCE SERVICES MODELS. SOURCE: LIMBA 2011.

Model Features Stages Models “E-Diamond” Model

1. Possible levels of implementation

There are 4 most common levels of implementation1.

The level of implementation is defined according to three different features – poling2.

2. Description of features for different levels

1) Every higher level includes all features of the lower level (stage) and complements them. 2) The first level deals with information publicity, whereas the highest level has a complete organizational integration.

1) The sevices are defined as a combination being individualized, general and performative of a certain level. 2) Every pole distributes the services from simple to more complicated ones.

2. Targeting at the client

Every higher stage integrates the clients even more, but the model is not suitable for every day life situations.

One of the poles is the relation between individualized and general.

3. Targeting at inner processes

The attention is paid. The fourth stage most commonly is devoted to interorganizational processes.

The model partly distinguishes targeting at inner processes.

4. Feedback The model does not envisage such an opportunity.

The model does not envisage such an opportunity.

5. Possibility for service assessment

It is available. It is assessed according to the level the service belongs to.

It is available. It can be assessed according to the place the service is in the “E-Diamond”.

6. Technological background

Every higher stage requires more modern technologies and better integration.

The more complicated the services of each pole are, the more complicated technological solutions they require in order to provide them.



www.elvedit.com 8

1. An exception is „Hiller & Belanger“ model which has 5 levels of implementation. 2. Service poles: separate coordinated; general individualized; informative

performative.

In evaluation of these general features designed for the models exceptionally from the point of traditional public administration conceptions, the most important of them are to be considered two of them – orientation towards the client and orientation towards internal organizational processes. However, the science of public administration, having assessed the information technology actively penetrating into all public administration system reform processes and public life, in this context becomes interdisciplinary; therefore, importance of other aforementioned general features of electronic government services models does not lose significance and value. After carrying out the comparative analysis of models, it is possible to make a presumption that both electronic government services stages’ model and the “E-Diamond” model are sufficiently proportionally oriented by the features of the orientation towards a client and orientation towards internal organizational processes.

Having carried out a comparative analysis of the models, it would be purposeful to analyze some similarities and differences among the models, identifying the causes and essentials of them.

The most simple comparison to carry out would be of the Stages Model and the “E-Diamond”. The “E-Diamond” model is made up of regrouping and complementing the stages of “SAFAD” Model into certain opposition. As the authors of the model claim [16], the first two stages of the “SAFAD” Model, namely the opportunities for information provision and ensurance of interaction in the electronic space, might be treated as the oppositin of informative and performative services, with regards to the services being used by the customer (see Picture 9). The third stage of the “SAFAD” Model implements the transactional services – receiving and transmitting information or electronic governance services. Services of a transactional nature are different from the other services in the level of individualization, thus in the authors’ opinion [16], the emergence of the opposition of individual and general electronic governance services would be purposeful. The fourth stage of the Stages Model, i.e. the Integration becomes an opposition to individualized and integrated services of the “E-Diamond” Model (Figure 2).



www.elvedit.com 9

FIGURE II: TRANSFORMATION OF THE STAGES MODEL INTO THE “E-DIAMOND” MODEL. SOURCE: LIMBA, 2011

It is worth paying attention to the fact that only the Stages Model provides the opportunity to assess the implementation of electronic governance services according to the levels set in advance. However, with the help of other models it can be identified what is the status of the electronic governance services provision. For instance, if the municipality provides only individualized, informative and general services, it might be assumed that either there is no technological basis or the integration of services having more complicated implementation (individualized, merged, implemented on the web) is avoided.

It is also possible to state that the designed “E-Diamond” model as it liberates the stages from their technological dependence, thus, it is simpler and more universal from the application aspect and in conceptual approach its implementation should be simpler. However, it is more likely that the main problem is not technological dependence of the stages’ models or issues of stages liberalisation while transforming them into the “E-Diamond” model, but the lack of competence by the officials of public administration institutions in the area of electronic government services project implementation [26]. The structure of the “E-Diamond” model itself is more complex than that of the stages’ models; thus, it could be more difficult to be understood and comprehended by civil servants responsible for the implementation of the projects of such nature in public sector. Consequently, the appearance of new models unambiguously requires also the changes in qualification and competence of civil servants.

V. CONCEPTUAL-HOLISTIC PROSPECTS OF IMPROVING ELECTRONIC GOVERNANCE SERVICES MODELS

It is difficult to distinguish one model that would be dominating or the best. Each model has its advantages and disadvantages. However, it is necessary to emphasize that these models are developed not only basing on theoretical paradigms, but also on practice of integrating electronic governance service into public sector in foreign countries. A great number of theoretical models,

Separate

General

Transactions

Transactions

Integration

Information

Coordinated

Individualized

Permormative

Interaction

Interaction

Informative



www.elvedit.com 10

such as ANAO, SAFAD, Lee & Layne are developed basing on experience of electronic governance service implementation in public administration institutions in Australia, Sweden and the USA. It should not be forgotten that all these models have been implemented and applied considering the peculiarities of public administration in every country, especially the specific system of local authorities and the national users needs. At this point it should be mentioned that the needs of local authorities systems and national users in different countries around the world vary, therefore claiming that one of the analyzed and compared models would suit perfectly to other countries, e.g. Lithuania, would be wrong [25]. Thus, we can draw a conclusion, that basing on the background of the above analyzed models, it would be purposeful to develop more universal electronic governance service models that would include holistic-social, competence and technological aspects, suitable for most countries public sector systems.

Some worldwide and European Union public administration institutions, especially municipalities, have a relatively low connection with local community while carrying out interactions. It is seen while poviding information, especially public services to the public. Having designed websites for the municipalities [1], it was thought that putting public information to electronic space would lead to more favourable conditions for the customers to use them, which would eventually solve the problem of the flow of residents applying directly to municipalities. However, some scientists [2, 6, 8, 9] tend to have doubts concerning that being the only solution. In such a case quite an important issue is identified and a question is asked directly relating to it – how could it be possible to provide services more efficiently and increase the flow of residents applying directly for public information and services while applying and developing the electronic governance service models. What is more, how can the activity and the importance of municipalities be fostered in the context of modernization.

In order to solve this problem, it is suggested to base on statements by Swiss scientists N. Thom and A. Ritz who state management of public institutions as well as municipalities demand new strategies, administrative and technological solutions from the fast changing environment [31]. However, existing legal conditions cast doubt on the reality of such a prospect.

What especially needs to be placed some emphasis on is the fact that those changes are resisted by administrative staff who are accustomed to the existing stable systems and are not inclined to innovations. The fact how public institutions are able to accept changes in the system and its environment and realize the factors that influence them has a considerable importance on public sector management.

Discretion of management differs in making influence on outside environment, human factors and conditions for institutions. The management’s role in influencing the conditions for political level is much lower than conditions for institutional level and human factors. The law on public organizations and public service still restricts flexibility, which is of extreme importance for the development of the modern reforms [31].



www.elvedit.com 11

It should not be forgotten, that public sector conditions are divided into: Outside conditions; Inside conditions.

The public sector outside conditions include only general results and results of specific influence

on outside environment, e.g. when there is provided final integrated and prepared public service for the user and additional consultancy service of how to use the implemented public service. The outside conditions provide the opportunity for the user to judge, e.g. if the public service is provided in a sufficiently qualified and effective manner [31]. Whereas inside conditions, that are affecting the management of public institutions, fall into the following:

Institutional conditions; Human factors.

Institutional conditions are targeted at selecting proper administrative structure and the

required number of employees, as well as the process of the executed reform. Usually the conditions of institutions are obvious and they set a clearly defined operation area for the management. Thus, while modernizing the administration, the pursue to reform and change those aspects is usually more reasonable than the pursue to change human aspect. However, the human aspects influence even more important organizational decisions, that are crucial to the proximity of outside factors as well, namely the directiveness of the public sector functions, depth and quality, the civil servants overcoming the barriers of social experience, knowledge and skills [31].

While confirming the above stated, we can base upon the presumption which is still relevant nowadays that was made by the USA scientist Kenneth L. Kraemer who has been exploring the peculiarities of the innovative tools implementation in the public sector. The author claims that one of the more important factors that limits the use of information technologies and the implementation of similar innovations in public sector is the lack of computer literacy and training among employees of public administration institutions. The use of information technologies provides more flexibility for such public sector organizations as municipality and its divisions [10, 19].

Having assessed the above mentioned scientific statements, it can be established that while implementing innovative means in local authorities institutions a great deal of attention must first of all be paid to the law reform, which is related to administrative changes due to the influence of innovations that are being implemented, and to the civil servants expertise development in the area of electronic governance service implementation. The emphasis must be placed on the fact that project implementation of electronic governance service and management in the local authorities level depend on the expertise and computer literacy of employees in municipalities. Electronic governance services first of all are implemented into the information-technological system of local authorities institutions and only then they can be provided to users. Thus, it can be stated that the quality and efficiency of providing electronic governance service



www.elvedit.com 12

depend on the level of knowledge and abilities of civil servants to use information technologies, as well as the outside and new instruments of public consultancy in the area of electronic governance services. Therefore, a great amount of attention must be drawn to changes in competence while designing and developing the new model of integrating electronic governance service, which would include holistic-social, competence and technological aspects.

VI. DESIGNING AND APPLICATION OF HOLISTIC ELECTRONIC GOVERNMENT SERVICES INTEGRATION MODEL FOR LOCAL AUTHORITIES

In discussing the aforementioned problem solving methods, various proposals are possible; however, one of them would be, particularly taking into account the problem of interaction between the institutions of district self-government and residents as well as business, – to create a model based on the principle of expert consultation application. External framework of the model would include the “Virtual Union of Local Authorities”. That is to say, it would be possible to merge local authorities into the single virtual unit on the cyberspace leaving their former usual structural functioning and the possibility to use the websites of these local authorities. The “Virtual Union of Local Authorities” model would be characterized by the implementation of the expert advisory function, e.g. for district municipalities, and this in its own turn would contribute to solving the issues of active use of electronic government services and their provision quality. To put it simply, the example could be a special model grounded by online expert consultations individually applied by one or several smaller district municipalities of Lithuania characteristic of the lowest activeness among residents addressing them as well as the larger city municipalities.

It is important to note that functioning of the “Virtual Union of Local Authorities” might be exceptionally virtual, on the contrary to other associations or public sector institutions. The “Virtual Union of Local Authorities” would function solely on the cyberspace or, more precisely, would exceptionally be characteristic of just the advisory function striving for improvement of interaction between local authorities and the public. Of course, attempting to consolidate such novelties in practice, though the advisory functioning of the proposed model, it would be worth to regulate it by legal acts of the Republic of Lithuania.

In considering the possibilities to apply the “E-Diamond” model in local authorities of Lithuania and having assessed the current situation of electronic government services integration, the development of the “lower edge of the diamond” fragments is possible. The lower edge of the “E-Diamond” model includes provision of services of more informative nature; however, this edge is integrated not gradually (one following another), but individually. In other words, all the edges of the model are independent of each other, in contrast to the models of electronic government stages. As in Lithuania the principles of stage models are known and applied, it is possible to state that in Lithuania practical application of the “E-Diamond” model has not been identified [24].



www.elvedit.com 13

In successful improvement of electronic government services introduction in local authorities, in future it would possible to discuss the development of the “E-Diamond” model in the upper “edge” fragments or the entire “E-Diamond” model integration.

However, having assessed quite poor experience of Lithuania in improving electronic government services, relevant electronic government service integration at self-government level problems, absence of alternative contemporary electronic government service models, the proposal would be to introduce the principle of “online expert consultation services” at the levels of “front-office” and “back-office” moving from the current electronic government stages’ model to the “E-Diamond” model, in this way designing a new and one of the most appropriate alternatives of Holistic Electronic Government Services Integration Model (see Figure 3).

Source: Compiled by the author FIGURE III: HOLISTIC ELECTRONIC GOVERNMENT SERVICES INTEGRATION MODEL

„Bac

k-of

fice

“

„Front-office“

Separate services

Informative services

General services

Permormative services

Individualizedservices

Coordinated services (Aligned

and Fused)

Hig

h co

mpe

tenc

e Local authority officials online

consulting services for citizens and

business (NGO) M

ore

effe

ctiv

e in

tegr

atio

n of

ele

ctro

nic

gove

rnm

ent

serv

ices

Online consulting services for

Local authority officials

Competitive changes of local

authority’s officials

MORE QUALITATIVE AND EFFECTIVE PROVISION OF

ELECTRONIC GOVERNMENT SERVICES

Knowledge sharing with other officials inside the

local authority

Knowledge sharing with other officials outside

organization with other municipalities

Low

com

pete

nce

Mor

e ef

fect

ive

cons

ulti

ng o

f cit

izen

s in

th

e sp

here

of e

lect

roni

c go

vern

men

t ser

vice

s

Stage model changing into “E-Diamond” (Goldkuhl,

G., Persson, A., 2006) model

MORE QUALITATIVE AND EFFECTIVE PROVISION OF

ELECTRONIC GOVERNMENT SERVICES

Saving of organization’s time and funds,

workload reduction of local authority’s officials

More efficient use of private and NGO organizations time

and funds

Influence of computer literacy of local

authority’s officials

Influence of computer literacy of citizens,

(clients, users)



www.elvedit.com 15

The first aspect “Low Competence” of this model structure is taken as a starting point showing the low organizational competence of a municipality in the area of electronic government services. What is more, it is important to point out that the implementation of the principle of online expert consultation services is closely related to and is dependent on the computer literacy skills of municipalities’ officials. Obviously, the higher the level of computer literacy of local authorities’ officials, the more effectively the online expert consultation services can be integrated in an organization, and vice versa. Thus, it is important to point out that the officials’ ability to participate in providing them with online expert consultation services is directly related to their own level of computer literacy, which is an important structural component of the aforementioned model.

At the “back-office” of the municipality a structural element “Qualification changes of municipalities’ officials” is created after the provision of the online expert consultation services. This element has two branches:

• Sharing the gained knowledge in the area of electronic government services within the organization, passing this knowledge on to all main structural levels of the organization;

• Sharing the gained knowledge in the area of electronic government services among organizations, spread of this knowledge among all local authorities (and other public administration institutions of the state) and their different structural levels.

As far as the holistic model of integrating electronic government services is concerned, the outcome aspect of the organizational inner structure of the municipality, which outlines the obtained result while implementing the principle of the online expert consultation, is a high organizational competence in the area of electronic government services. However, the main structural element of the outcome of this model that helps to withdraw inefficient stage model and move on to the three alternate poles (six elements) implementation of the “E-Diamond” model is the more efficient integration of electronic government services, which ensures saving the costs of the municipality budget and time as well as the reduction of the workload of municipalities’ officials more effectively.

At the “back-office” level the online expert consultation services, as it has already been mentioned, would firstly be provided for the officials of municipalities. While at the “front-office” level of an organization having become qualified specialists of this area the officials of municipalities providing online expert consultation services for the public could inform residents and businesspeople about the electronic government services and their advantages. It has to be mentioned that at the “front-office” level the online expert consultation services in the area of electronic government services would be provided for residents and business entities on the principle of online consultation. Of course, here other, quite important problems of the public motivation to participate in seminars of such nature and computer literacy arise, on which consultation effectiveness, saving finances and time of the interested, and finally, key expected result – a more high-quality and efficient electronic government service provision – depend.

The “online expert consultation service level” principle is characteristic of universality, therefore, it can be quite easily integrated into the aforementioned and on this basis designed



www.elvedit.com 16

Holistic Electronic Government Services Integration Model. “Online expert consultation service level” principle can be easily implemented also in practice.

Based on the proportion of electronic government service model comparison by orientation into municipal internal processes and orientation towards satisfaction of client needs, currently the most appropriate to Lithuania and forming the conditions for the occurrence of the new research possibilities would be suggested and created the Holistic Electronic Government Services Integration Model on “E-Diamond” model basis. This Holistic Electronic Government Services Integration Model was also designed using the basis of the survey of the electronic government services provision situation in Lithuania (Limba, T, 2009, B).

It is presumable that after application and implementation of the proposed Holistic Electronic Government Services Integration Model, more favourable mutual conditions would be formed: to municipalities – to more effectively integrate electronic government services, and to the public – to more effectively and in a more high-quality manner make use of electronic government services provided by both larger (city) and smaller (of district) municipalities.

The practical application realization of Holistic Electronic Government Services Integration Model with the design of “The project of implementing the principle of online expert consultation on the model of electronic government services” could be created and analyzed in detail for the future investigations.

VII. RESEARCH OF APPLICATION HOLISTIC ELECTRONIC GOVERNMENT SERVICES INTEGRATION MODEL IN LITHUANIAN MUNICIPALITIES

A. Experts‘ evaluation methodology and data analysis

Qualitative research method had been applied by questioning various Lithuanian authorities and institutions experts, whose work relates to the implementation of electronic government services. The aim of experts’ evaluation – to identify the importance and relevance of instrumentation of Lithuanian residents’ and municipalities’ officials research. Lithuanian residents’ research was carried out using 27 questions questionnaire. Municipalities’ officials were asked to answer 28 questions. All were asked to indicate their gender, age, place of residence, education, employment, office, computer literacy, to provide answers to questions related to an evaluation of electronic government services. The questionnaire consists of the questions in accordance with the principles of the questionnaire formation. Before submitting the questionnaire for respondents, they were aware of the purpose, relevance of problem. There has been shown that the form is anonymous and the data will be used generally. Questionnaire identifying the key explanations and instructions on how to fill it.

Experts evaluated questionnaires for Lithuanian residents and municipalities’ officials and ranked the questions according to the importance. Based on the foregoing method, nine Lithuanian experts were interviewed. Lithuanian experts were interviewed about two questionnaires to residents and municipalities’ officials without the socio-demographic characteristics of reflective questions, only those issues that have a close link to electronic



www.elvedit.com 17

government services were evaluated. After evaluation, the least important questions were taken away from the questionnaires.

B. Residents‘ research methodology

Based on quantitative research methods the questionnaire survey method was used in order to obtain the opinion of the residents of Lithuania about Lithuanian municipalities offered electronic government services and their adaptation capabilities.

The purpose of residents‘ research – to investigate the population favor evaluating electronic government services adaptation capabilities in Lithuanian municipalities.

There have been introduced 2 hypotheses:

1. Respondents’ quality evaluation of electronic government services in municipalities depends on their own ability to use electronic government services;

2. Respondents’ ability to access electronic government services depends on the competence of civil servants to advice in this area.

The dependence between the two variables to identify and test hypotheses to confirm or deny, the calculation of Spearman's correlation coefficient used by “STATISTICA” data processing program.

Sampling Method - by giving questionnaires for residents, respondents were selected through probability samples nested sampling method. Questionnaires were proportional to the number of total population of the municipality [29]. Required sample calculated using http://www.surveysystem.com/sscalc.htm website. Calculation of the sample indicated that the confidence level is 95 percent, 5 percent confidence interval, population – 2,7 million. The estimated sample – 384. There had been surveyed 420 respondents. Distributed 458 questionnaires. Distributed questionnaires have 81 percent return rate.

C. Residents‘ research data analysis

Residents were asked to evaluate municipalities' officials ability to consult in electronic government services area. 8 percent of surveyed respondents municipalities’ officials ability to consult in electronic government services area evaluated 5 points out of ten. Equal parts – 5 percent of respondents municipalities’ officials ability to consult in electronic government services area evaluated 6 and 7 points. 4 percent of respondents municipalities’ officials ability to consult in electronic government services area evaluated 8 points out of ten. Only 2 percent of respondents evaluated municipalities’ officials ability to consult in electronic government services area with the highest scores - 9, 10. The rest of the respondents ability to consult in electronic government services area evaluated below 2 percent or did not answer to this question (71 percent) (see Figure 4).



www.elvedit.com 18

FIGURE IV: EVALUATION OF MUNICIPALITIES’ OFFICIALS ABILITY TO CONSULT IN ELECTRONIC GOVERNMENT SERVICES AREA

Also, respondents expressed their opinion about expert consultation necessity for

municipalities’ officials in electronic government services area. Almost half of surveyed respondents - 42 percent believes, that municipalities’ officials need expert advice in electronic government services area, so that they could later be precisely introduced and significantly encourage residents and business to use electronic government services. 30 percent of respondents answered that municipalities’ officials need expert advice in electronic government services area, but it would have a slight impact on the population using electronic government services. 5 percent of respondents think that it has absolutely no influence on more active citizens and businesses using electronic government services. 7 percent of respondents believe that municipalities’ officials do not need expert consultations about electronic government services, because more active people and businesses using electronic government services do not depend on them. 12 percent of respondents have no opinion on this issue (see Figure 5).



www.elvedit.com 19

FIGURE V: RESPONDENTS’ OPINION ABOUT EXPERT CONSULTATION NESCESSITY FOR MUNICIPALITIES’ OFFICIALS IN ELECTRONIC GOVERNMENT SERVICES AREA

After analyzing the data, it was important to approve or to deny hypothesis.

1 hypothesis. Respondents’ quality evaluation of electronic government services in municipalities depends on their own ability to use electronic government services:

Spearman R - Spearman ordinal correlation coefficient is 0,624282 - medium positive correlation: we can assume that when the respondent evaluates the quality of electronic government services better, then he evaluates his ability to use electronic government services better.

Formulating H0 and the alternative hypothesis Ha: H0 - the correlation coefficient is equal to zero, or the relationship between the variables is not available, the Ha - the correlation coefficient is not zero, then the relationship between the variables exists.

Selecting the significance level α = 0,05 (five percent error), in social studies, it is recommended choose this error.

p-level - observational significance level (p-level = 0,00 <α = 0,05) refers to prove alternative hypothesis Ha and Spearman correlation coefficient is significantly different from zero. Hence, respondents electronic government service quality depends on their own ability to use electronic government services.

2 hypothesis. Respondents’ ability to access electronic government services depends on the competence of civil servants to advice in this area:

Spearman R - Spearman ordinal correlation coefficient is 0,434538 - a weak positive correlation close to the average: we can assume that when the respondent evaluate the competence of civil servants better, then he evaluates his ability to use electronic government services better.



www.elvedit.com 20



p-level - observational significance level (p-level = 0,00000 <α = 0,05) refers to prove alternative hypothesis Ha and Spearman correlation coefficient is significantly different from zero. Respondents’ ability to access electronic government services depends on the competence of civil servants to advice in this area.

The results of residents’ research show that the interest in using electronic government services is large enough, but the level of computer literacy, affecting the ability to use the same services is quite low. Another very important reason to sway public skills and competence in the use of electronic government services - a poor Lithuanian municipal public awareness and familiarization with modern facilities, saving time and money to get public services. Research results showed, that according to the majority of respondents, the use of public services activity would increase, it would save time and financial costs, if Lithuanian municipalities use more electronic government services.

Based on the research results, it can be stated that the municipalities’ officials lack of competence in the field of electronic government services is one of the major reasons causing the lack of electronic government services prevalence and use among the public. Research results revealed problems of electronic government services integration in municipalities and demonstrate the need of implementing Holistic Electronic Government Services Integration Model. However, only residents’ survey to approve the above mentioned model implementation in Lithuanian municipalities system is not enough. In this case, it is also important to carry out a research of municipalities’ officials, which data analysis would help to evaluate the need of this model on both sides.

D. Municipalities’ officials research methodology

Based on qualitative research methods questionnaire survey method was also applied in order to obtain Lithuanian municipalities’ officials opinion about Lithuanian municipalities offered electronic government services and their adaptation capabilities.

The purpose of municipalities’ officials research – to determine, how Lithuanian municipalities’ officials evaluate their institutions provided electronic government services adaptation options.

There have been introduced 4 hypotheses: 1. The municipal budget savings depend on direct appeal flow to the municipality

associated with the electronic government services rendering to the public. 2. Municipalities’ officials work load changes depend on direct appeal flow to the

municipality associated with the electronic government services rendering to the public.



www.elvedit.com 21

3. Municipalities’ officials time costs change depends on direct appeal flow to the municipality associated with the electronic government services rendering to the public.

4. Municipality provided electronic government services quality evaluation depends on the respondents' knowledge evaluation in the area of electronic government services.

Submitting questionnaires to municipalities’ officials, the respondents were selected through probability sampling, random sampling method. Required sample calculated using http://www.surveysystem.com/sscalc.htm website. Calculation of the sample indicated that the confidence level is 95 percent, confidence interval is 3 percent, population – 7500. The estimated sample – 934. There had been surveyed 1301 respondents. Distributed questionnaires have 85 percent return rate.

E. Municipalities’ officials research data analysis

It was important to find out municipalities' officials familiarity with public services transfering to internet maturity levels according to the Lithuanian E-government concept. 62 percent of respondents said, that they heard nothing about public services transfering to internet maturity levels according to the Lithuanian E-government concept. 25 percent of respondents accidentally found out about public services transfering to internet maturity levels according to the Lithuanian E-government concept from other sources of information. 9 percent of respondents noted that they are personally interesting in public services transfering to internet maturity levels according to the Lithuanian E-government concept. 4 percent of respondents indicated that they learned about public services transfering to internet maturity levels according to the Lithuanian E-government concept in special consultative lecture-seminars (see Figure 6).

FIGURE VI: MUNICIPALITIES' OFFICIALS FAMILIARITY WITH PUBLIC SERVICES TRANSFERING TO INTERNET MATURITY LEVELS ACCORDING TO THE LITHUANIAN E-GOVERNMENT CONCEPT

Also, municipalities’ officials were asked to evaluate their own knowledge associated with

consulting in electronic government services area. Equal parts - 9 percent of surveyed respondents rated their ability to advise in electronic government services area 5, 7, and 8



www.elvedit.com 22

points out of ten. 7 percent of respondents rated their ability to advice in electronic government services area 6 points out of ten. Equal parts - the 4 percent of surveyed respondents rated their ability to advice in electronic government services area 3, 4 and 9 points. Only 2 percent of respondents rated their ability to consult in electronic government services area with the highest grade – 10. 49% of respondents had never consulted in electronic government services area (see Figure 7).

FIGURE VII: EVALUATION OF MUNICIPALITIES’ OFFICIALS THEIR OWN KNOWLEDGE ASSOCIATED WITH CONSULTING IN ELECTRONIC GOVERNMENT SERVICES AREA

After analyzing the data, it was important to approve or evaluate their own knowledge

associated with consulting in electronic government services area to deny hypothesis.

1 hypothesis. The municipal budget savings depends on direct appeal flow to the municipality associated with the electronic government services rendering to the public:

Spearman R - Spearman ordinal correlation coefficient is 0,434538 - a weak positive correlation close to the average: we can assume that the more decrease direct referral flow to the municipality associated with electronic government services rendering to the public, the more the municipality saves budget.



p-level - observational significance level (p-level = 0,00 <α = 0,05) refers to prove alternative hypothesis Ha and Spearman correlation coefficient is significantly different from zero. The municipal budget savings depend on direct appeal flow to the



www.elvedit.com 23

municipality associated with the electronic government services rendering to the public.

2 hypothesis. Municipalities’ officials work load changes depend on direct appeal flow to the municipality associated with the electronic government services rendering to the public:

Spearman R - Spearman ordinal correlation coefficient is 0,406339 - a weak positive correlation: we can assume that the more decrease direct referral flow to the municipality associated with electronic government services rendering to the public, the more reduces the municipalities’ officials workload.



p-level - observational significance level (p-level = 0,00000 <α = 0,05) refers to prove alternative hypothesis Ha and Spearman correlation coefficient is significantly different from zero. Municipalities’ officials workload changes depend on direct appeal flow to the municipality associated with the electronic government services rendering to the public.

3 hypothesis. Municipalities’ officials time costs change depends on direct appeal flow to the municipality associated with the electronic government services rendering to the public:

Spearman R - Spearman ordinal correlation coefficient is 0,339406 - a weak positive correlation: we can assume that the more decrease direct referral flow to the municipality associated with electronic government services rendering to the public, the more reduce the municipalities’ officials time costs.



p-level - observational significance level (p-level = 0,00000 <α = 0,05) refers to prove alternative hypothesis Ha and Spearman correlation coefficient is significantly different from zero. Thus, municipalities’ officials time costs change depends on direct appeal flow to the municipality associated with the electronic government services rendering to the public.

4 hypothesis. Municipality provided electronic government services quality evaluation depends on the respondents' knowledge evaluation in the area of electronic government services:

Spearman R - Spearman ordinal correlation coefficient is 0,350651 - a weak positive correlation: we can assume that when the respondent evaluates his knowledge in the field of electronic government services better, then he evaluates the quality of municipality provided electronic government services better.



www.elvedit.com 24



p-level - observational significance level (p-level = 0,00 <α = 0,05) refers to prove alternative hypothesis Ha and Spearman correlation coefficient is significantly different from zero. Consequently, municipality provided electronic government services quality evaluation depends on the respondents' knowledge evaluation in the area of electronic government services.

Research results showed that according to the majority of respondents, the use of public service activity would increase, it would save time and financial costs, if Lithuanian municipalities use more electronic government services. It has been determined that the majority of municipalities’ officials are not familiar with the models of electronic government services. It is important to note that the majority of municipalities’ officials are not familiar with current and identified electronic government services stage model in Lithuania.

Based on the results of the research, it could be said that municipalities’ officials experience in advising via Internet (together - by e-mail, and "online" system) about the availability of municipality services offered by electronic government services, are sufficiently low. That notes and the data that only 1,45 percent of municipalities’ officials advised to residents or private entities about using electronic government services online. It can be said that the municipalities’ officials qualification on consulting online about electronic government services should be improved.

Municipalities’ officials research results revealed problems of electronic government services integration in municipalities and demonstrate the need of implementing Holistic Electronic Government Services Integration Model. It stimulate to conduct the experiment of implementing this model in Lithuanian municipalities.

F. Experiment methodology

Created Holistic Electronic Government Services Integration Model performance based on an experiment involving Lithuania municipalities’ officials. Experiment was treated in accordance with the quantitative research methods.

The purpose of the experiment - to determine the implementation and action possibilities of created electronic government services model in Lithuanian municipalities. There have been set the following objectives for the above mentioned purpose to be achieved:

1. To hold networked (online) expert consultation seminars for municipalities’ officials; 2. To familiarize municipalities’ officials with electronic government services problems

and performance contexts, electronic government services models through online expert consultation seminars, presented scientific material.

The hypothesis – quality of Lithuanian municipalities’ officials in expert “online” seminars given tasks depends on duration of tasks.



www.elvedit.com 25

Sampling Method - in the experiment, the respondents were selected through non-probability sampling, occasional sampling (when the sample included most conveniently analyzed elements). The study involved two respondents from each of the municipalities. Respondents from Šiauliai, Pakruojis, Anykščiai, Molėtai, Švenčionys and Druskininkai municipalities analyzing 4 subjects performed the 4 tasks in each theme, responded to a question in the first theme and two questions on rest topic (total - 3 themes). Respondents were analyzed scientific material related to electronic government services development aspects and features, models and application of these models, also coordination opportunities in Lithuanian local level. At the end of the research respondents answered two control questions intended to acquire the knowledge and skills inspection.

G. Experiment data analysis

It was important to compare, how much time respondents spent for the online expert consultation workshop tasks. Minimum time spent for the online expert consultation workshop tasks was 65 minutes, maximum – 413 minutes. Usually respondents spent from 150 to 250 minutes for the online expert consultation workshop tasks (see Figure 8).

FIGURE VIII: COMPARISON OF SPENT TIME FOR THE ONLINE EXPERT CONSULTATION WORKSHOP TASKS

Also, it was important to define quality criterias total evaluation assessment comparison of respondents’ given tasks. There were nine tasks (questions) and for every task respondent could get 1 point. Maximum he could get 9 points for all tasks. Every task was evaluated according to these criterias:

Criterion 1: “Respondent understood the problem exactly” - evaluation of 1 point; Criterion 2: “Respondent realized the problem approximately” - evaluation of 0,75

points; Criterion 3: “Respondent partially realized the problem” - evaluation of 0,5 points; Criterion 4: “Respondent had difficulty in understanding the problem” - evaluation of

0,25 points; Criterion 5: “Respondent did not understand the problem” - evaluation of 0 points.



www.elvedit.com 26

The lowest evaluation was 7,5 points, one respondent got maximum of evaluation – 9 point.Most of respondents were evaluated 8, 7,5 or 7,75 points (see Figure 9).

FIGURE IX: QUALITY CRITERIAS TOTAL EVALUATION ASSESSMENT COMPARISON OF RESPONDENTS’ GIVEN TASKS

After the Holistic Electronic Government Services Integration Model application in Lithuanian municipalities experiment, it was important to approve or to deny hypothesis. In order to determine a single variable (X - tasks quality) dependence on another variable (Y - task completion time), the linear regression equation need to be calculated. Calculated using the following equation derived value of the coefficient of determination R2 = 0,8816. In order to determine the dependence of one variable on another variable strength, it is necessary to calculate the Pearson correlation value. Pearson correlation coefficient value can be calculated using the coefficient of determination and is expressed in the following formula:

r =√ R2

The estimated correlation coefficient value obtained r = 0,9389 indicates that the variable interdependence relationship is strong. The hypothesis can be approved – quality of Lithuanian municipalities’ officials in expert “online” seminars given tasks depends on duration of tasks. In other words, it can be said that the longer a respondent takes to perform this task, the higher the quality of tasks and vice versa - the less time the respondent spends, the worse the quality of task performance.

Respondents of experiment noted that they gained additional knowledge of electronic government services development opportunities for municipalities, especially they gained knowledge about implementation of electronic government services models at the municipal level.

Respondents at the end of the control questions indicated that knowledge are going to be used in practice, for example, designing and developing project of e-Šiauliai region. Respondents on their own initiative also noted that they will recommend for others



www.elvedit.com 27

municipalities’ officials and professionals to participate in this "online" consultation seminar intended for municipalities’ officials to improve qualification of electronic government services and strive to raise public awareness of electronic government services more effective integration and use.

In summary, it can be stated that the current situation of electronic government services, residents and municipalities’ officials assessment analysis and the results of the experiment carried out, confirm that the Holistic Electronic Government Services Integration Model is appropriate, can be effectively integrated and applied in Lithuanian local authorities.

VIII. CONCLUSION AND RECOMMENDATIONS 1. Having carried out comparative analysis of stage and “E-Diamond” models of

electronic government services, there are distinguished six key features of the models, such as possible levels of implementation, attributes of different levels, targeting at the customer, targeting at the inside processes, feedback, possibility to evaluate services, technological background. However, out of six features the main ones are considered to be the feature of targeting at the inside organizational processes of self-government and the feature of targeting at the customer. The latter is emphasized most of all due to the fact that it represents customers interests best.

2. The implementation of the above mentioned features is identified in the models of “Stages” and “E-Diamond”. Both of them are quite equally targeted at restructuring inside processes and meeting the needs of customers. Assessing the models of “Stages” and “E-Diamond” according to this rather neutral targeting, they can be applied in economically developing as well as highly developed countries, thus can be considered to be universal.

3. Aiming at a greater universality and practical applicability of models, the perspective of model improvement should be oriented towards improvement of conceptual-holistic processes under the external and internal conditions of the public sector system. The significance of internal conditions of public sector, first of all, is to be linked with human resources management peculiarities, upgrading of their competence and qualification skills. Of course, here a great role is played also by computer technology. Thus, in this case it is essential to emphasize that application and management of specific electronic government service models at local self-government level depend on the overall holistic processes – competence of municipal officials in innovation management area, their conceptual abilities in electronic government knowledge, application and proper formation of technology skills and computer literacy. Thus, it is possible to formulate a conclusion that quality and effectiveness of electronic government services provision to consumers depend on the knowledge of civil servants, level of their ability to use information technology, external and new public consultation in electronic government service area instruments.

4. Despite the fact that in the old member states of the European Union – Austria, Netherlands, Belgium, Germany the principles of the electronic government administration are becoming a norm, having stepped through the threshold of the twentieth century, in the central and local authorities of the other European states, for example Lithuania, there is still widely applied the Max Weber’s model of hierarchical bureaucracy, which impedes the success o innovations as well as the implementations of principles and models of electronic



www.elvedit.com 28

government services. Therefore, to enable the establishment of the suggested Holistic Electronic Government Services Integration Model in Lithuanian municipalities, it is not sufficient to solve its application and implementation problems causing only competitive, managerial organizational changes. Practical implementation of the models of electronic government services in Lithuania should also be regulated by legal acts.

5. Provision of public services for the society is one of the realized and regularly developing functions of municipalities worldwide. The range of the services for residents and business is rather big, thus awareness of electronic government services provided by municipalities would be critical for nearly all levels of municipalities’ officials. Consultations provided by experts during online consultation services based on the suggested Holistic Electronic Government Services Integration Model, that was created in accordance with “E-Diamond” model, could help municipalities’ officials get accustomed with features of providing electronic government services to residents and business entities, as well as could provide an opportunity to become more competitive and efficient specialists in this area. What is more, the principle of online expert consultation of municipalities’ officials of the suggested model could be implemented in the “back office” or/and “front office” of the public sector in Lithuania. Finally, it can be claimed that having implemented and widely applied the suggested Holistic Electronic Government Services Integration Model, more efficient integration of electronic government services in the local self-government level and more accurate implementation of public expectations might be anticipated.

6. Research of application Holistic Electronic Government Services Integration Model in Lithuanian municipalities have been conducted. Methodological research applied expert survey method, various Lithuanian authorities and institutions experts, whose work relates to the implementation of electronic government services were interviewed. The questionnaire survey method was used in order to obtain the opinion of the residents of Lithuania about Lithuanian municipalities offered electronic government services and their adaptation capabilities. Questionnaire survey method was also applied in order to obtain Lithuanian municipalities’ officials opinion about Lithuanian municipalities offered electronic government services and their adaptation capabilities. Created Holistic Electronic Government Services Integration Model performance based on an experiment involving Lithuanian municipalities’ officials. It can be stated that the current situation of electronic government services, residents and municipalities’ officials assessment analysis and the results of the experiment carried out, confirm that the Holistic Electronic Government Services Integration Model is appropriate, can be effectively integrated and applied in Lithuanian local authorities.

REFERENCES [1] Ancarini, A. (2005). Towards quality e-service in the public sector: The evolution of web sites in the

local public service sector, Managing Service Quality, Vol. 15, 6-23.

[2] Andersen, K.V.; Henriksen, H.Z. (2003). E-government maturity models: extension of the Layne and Lee model. Government information Quaterly. No. 26, 236–248.

[3] Andersen, K. V. (2004). E-government and public sector process rebuilding. Amsterdam: Kluwer.



www.elvedit.com 29

[4] Becker, J., Algermissen, L., Niehaves, B. (2005). Processes in E-Government focus: A procedure model for process oriented reorganisation in public administration on the local level. Accepted to the First International Pragmatic Web Conference, September, 87–103.

[5] Becker, J., Algermissen, L., Niehaves, B. (2006). A procedure model for process oriented E-Government projects. Accepted to the First International Pragmatic Web Conference, September, 150–183.

[6] Bretschneider, S. (2003). Information technology, e-government, and institutional change. Public Administration Review, 63(6), 738–744.

[7] Buckley, J. (2003). E-service quality and the public sector, Managing Service Quality. Vol 13, 453-462.

[8] Burn, J., Robins, G. (2003). Moving towards e-government: A case study of organisational change processes. Logistics Information Management, 25–35.

[9] Coe, A., Paquet, G., Roy, J. (2001). E-Governance and Smart Communities – A Social Learning Challenge, Social Science Computer Review, 80-93.

[10] Davenport, T.H. (1999). Process Innovation: Reengineering Work through Information Technology, Boston, MA: Harvard Business School Press, 246-267.

[11] Domarkas, V., Lukoševičienė, V. (2006). Electronic government by the aspect of providing information for the society. Public policy and administration, No. 16, 73–86.

[12] Fountain, J. (2001). Building the virtual state: Information technology and institutional change. Washington, DC, Brookings Institution, 64-78.

[13] Goldkuhl, G. (2005). Socio-Instrumental Pragmatism: A Theoretical Synthesis for Pragmatic Conceptualisation in Information Systems, in Proceedings of the 3rd Intl Conf on Action in Language, Organisations and Information Systems, Limerick, 115-132.

[14] Goldkuhl, G., Cronholm, S., Sjostrom, J. (2004). User Interfaces as Organisational Action Media, in Proceedings of the 7th International Workshop on Organisational Semiotics (pp. 124-140). Portugal.

[15] Goldkuhl, G.; Persson, A. (2006). Characteristics of Public E-services: Investigating the “E-Diamond” Model. Accepted to the First International Pragmatic Web Conference, September (pp. 54-79). Stuttgart, Germany.

[16] Goldkuhl, G., Persson, A. (2006). From e-ladder to “E-Diamond” – re-conceptualising models for public e-services. Proceedings of the 14th European Conference on Information Systems (pp. 117-132). Göteborg.

[17] Gronlund, A. (2002). Electronic Government - Design, Applications and Management. Hershey et al.: Idea Group Publishing, 61- 77.

[18] Gugliota, A., Cabral, L., Doingue, J., Roberto, V., Rowlatt, M., Davies, R. (2005). A semantic web service-based architecture for the interoperability of e-government services, 133–145.

[19] Kraemer, K. L., King, J. L. (1996). Information technology and administrative reform: Will the time after e-government be different? Irvine, CA7 CRITO, University of California, 580-582.



www.elvedit.com 30

[20] Layne, K.; Lee, J. (2001). Developing fully functional e-government: a four stage model. Government information Quarterly. No. 18, 122–136.

[21] Lenk, K., Traunmueller, R. (2001). Broadening the Concept of Electronic Government. In: Prins, J. E. J. (Ed.) Designing E-Government. Amsterdam: Kluwer, 63-74.

[22] Lind, M., Forsgren, O., Salomonson, N., Albinsson, L. (2004). The E-co model – citizens‘ driving e-services quality, 97–124.

[23] Limba, T. (2009) Electronic government services’ maturity models: their comparative analysis. Information Sciences, 30-40.

[24] Limba, T. (2009). Models of electronic government services: opportunities of their application in Lithuanian municipalities. (Doctoral disertation, Mykolas Romeris University)

[25] Limba, T. (2011). Comparative analysis of Stages models and “E-Diamond” model of electronic government services, the conceptual features of their improvement. Information Sciences, 8-23.

[26] Limba, T. (2011) Comparative aspects of Electronic government services’ maturity and “E-Diamond” models. Social Technologies, 1(1), 35-45.

[27] Millard J. (2003). The (r)e-Balancing of Government. In e-Government: Public Administration for a New Century. UPGRADE. IV (2).

[28] Persson A., Goldkuhl G. (2005). Stage-models for public e-services – investigating conceptual foundations. 2nd Scandinavian Workshop on e-government, February, Copenhagen, 151–188.

[29] Rudzkienė, V. (2005) Social Statistics. Vilnius, MRU publishing center.

[30] Signore, O.; Chesi, F.; Pallotti, M. (2003). E-government: challenges and opportunities. CMG Italy – XIX Annual Conference, Florence, Italy, 231–265.

[31] Thom, N., Ritz A. (2004). Public management. Monography. Vilnius: LTU.

AUTHORS’ BIOGRAPHY

Tadas Limba was born in Vilnius, Lithuania in 1976. He got B. Sc. in Politics from Vilnius University, 1999 and B. Sc. in Law from Mykolas Romeris University, 2010. He got M. Sc. in Public Administration from Mykolas Romeris University, 2001 and M. Sc. in Business Law from Mykolas Romeris University, 2012. Also, Tadas Limba got his Ph. D. in Management and Administration from Mykolas Romeris University, 2009. Tadas Limba is an Associate Professor from 2010. Since 2012 he also is a head of Institute of Digital Technologies. He has published about 20 articles in several areas of information science,

monography, handbook, especially in areas of e-government and e-business. Tadas Limba is a member of Lithuanian Computer Society since 2007. Since 2013 he is Asia Center Board Member, South Korea's representative at Mykolas Romeris University. He plays an active role in international communication and development of joint double degree studies program with South Korea Dongseo University. Also, Tadas Limba made presentations at international and national conferences.



www.elvedit.com 31

Gintarė Gulevičiūtė was born in Panevėžys, Lithuania in 1989. She got B. Sc. in Public Administration from Mykolas Romeris University in 2008. She got M. Sc. in Electronic Business Management from Mykolas Romeris University. Now she is an assistant in Institute of Digital Technologies at Mykolas Romeris University. In 2013 she published an article about Peculiarities of Electronic Government Services Implementation in European Union. Her areas of interest is e-government and e-business. Gintarė Gulevičiūtė is the coordinator of Digital Content Academy at Mykolas

Romeris University. During her study years she has organized conference “Future business 2013“ at Mykolas Romeris University.


Detecting Suspicion Information on the Web Using Crime Data Mining Techniques

Authors

Javad Hosseinkhani Department of Computer Engineering, Islamic Azad University, Zahedan Branch, Iran

[email protected] Zahedan, Iran

Mohammad Koochakzaei Department of Computer, Science and Research Branch, Islamic Azad University, Zahedan, Iran


Solmaz Keikhaee Department of Electrical Engineering/ Islamic Azad University, Science and Research Branch, Iran

Yahya Hamedi Amin Department of Computer Engineering, Qeshm International Branch, Islamic Azad University, Iran


[email protected] Qeshm, Iran

Abstract

Along with the rapid popularity of the Internet, crime information on the web is becoming increasingly rampant, and the majority of them are in the form of text. Because a lot of crime information in documents is described through events, event-based semantic technology can be used to study the patterns and trends of web-oriented crimes. The purpose of this paper is to provide a review to mining useful information by means of Data Mining. The procedure of extracting knowledge and information from large set of data is data mining that applying artificial intelligence method to find unseen relationships of data. There is more study on data mining applications that attracted more researcher attention and one of the crucial field is criminology that applying in data mining which is utilized for identifying crime characteristics. Detecting and exploring crimes and investigating their relationship with criminals are involved in the analyzing crime process. Criminology is a suitable field for using data mining techniques that shows the high volume and the complexity of relationships between crime datasets. Therefore, for further analysis development, the identifying crime characteristic will be the first step and obtained knowledge from data mining approaches is a very useful tool to help and support police forces. This research aims to provide a review to extract

Detecting Suspicion Information on the Web using Crime Data Mining Techniques Javad Hosseinkhani, Mohammad Koochakzaei, Solmaz keikhaee, and Javid Hosseinkhani Naniz

Copyright © 2014 Helvetic Editions LTD - All Rights Reserved www.elvedit.com 33

useful information by means of Data Mining, in order to find crime hot spots out and predict crime trends for them using crime data mining techniques.

Key Words

Cyber Crime, Web Crime Mining, Crime Data Mining Techniques, Forensics Analysis, Web Mining

I. INTRODUCTION Cyber crimes mean that the illegal activities are committed through the use of computers and

the Internet. Cyber crimes can be basically divided into two major categories. One is those take the network as criminal objects such as trespassing, destructing the network system, etc. The others are those using the network to commit crime such as fraud, eroticism, illegal trade, etc.

The Internet has created fertile ground for cyber crimes. Information of violence, pornography, fraud can be seen everywhere on the Internet. According to a statistics report conducted by researchers from Taiwan area and Japan, the most common proportion of illegal network usage cases in sequence are: Internet pornography, Internet fraud, trafficking in illicit goods, intimidation and extortion, illegal intrusion, insult and slander.

A lot of facts have proved that it is not enough to manage the information on the Internet simply through traditional administrative models. In this concern, Web mining is a novel research direction for the information gathering and analyzing on the Internet, which is explosive and unstructured. The focuses of Web mining research are to develop new web mining techniques and to extract the features of texts to represent them.

Criminal web data always offer valuable and appropriate information for Law administration. The evaluation of the different capacities of widespread criminal web data is very difficult all the time so it is one of the most noteworthy tasks for law administration. Crimes may be as extreme as murder and rape where advanced analytical methods are required to extract useful information from the data Web mining comes in as a solution [1, 3].

Definitely, one of influential factors that encounter a crime phenomenon is the humans’ social life circumstances so the crime analysis knowledge is needed as an efficient combating tool. It also comprises of leveraging a systematic approach for discovering, identifying and predicting crime incidents and its input is contained assigned information and data in crime variables and the output contains the answer to knowledge extraction, analytical and investigative questions and the visualization of the results. Due to the criminality-related data and crime complexity and also the existence of intangible relations between them, data mining a rapidly made in growing field among criminologists. In the police departments, large volumes of crime-related data are existed. Due to the crime-related complexity relationships, the traditional methods of crime analysis are out-of-date that consume many time and human resources. Moreover, these methods are not able to obtain all influential parameters because of their high amount of human interference, therefore, using an intelligent and systematic approach for crime analysis more



www.elvedit.com 34

than ever. However, the data mining techniques can be the key solution [4].

Areas of concentrated crime are often referred to as hot spots. Researchers and police use the term in many different ways. Like researchers, crime analysts look for concentrations of individual events that might indicate a series of related crimes. They also look at small areas that have a great deal of crime or disorder, even though there may be no common offender. Analysts also observe neighborhoods and neighborhood clusters with high crime and disorder levels and try to link these to underlying social conditions[28].

Nowadays, the accessible data sources are provided by the rapid growth of the Web that has many specific characteristics. In fact, these characteristics make the mining useful knowledge and information a challenging task. It is necessary to know data mining in order to discover information mining on the Web that is exist in many Web mining tasks. Though, Web mining is not completely the application of data mining [2].

Data mining is defined as the process of discovering, extracting and analyzing meaningful patterns, structure, models, and rules from large quantities of data. Data mining is emerging as one of the tools for crime detection, clustering of crime location for finding crime hot spots, criminal profiling, predictions of crime trends and many other related applications [28].

The aim of web mining is to extract appropriate information from the page content, Web hyperlink structure and usage data. Although Web mining uses many data mining techniques, it is not purely an application of traditional data mining due to the heterogeneity and semi-structured or unstructured nature of the Web data [2].

The user is interested to identify crime hot spots of a particular region on certain crime types for a specific period. In order to fulfill such a requirement, a user interactive query interface is needed. Kumar et al [6], has presented an interactive media system with capability to adapt to various conditions from user preferences and terminal capabilities to network constraints. Newsome et al. [7] has proposed HyperSQL as a web-based query interfaces for biological databases. The design of query interfaces to biological database has also been presented by Che. et al. [8]. However, no online adaptive query interface has been designed for mining crime data. The purpose of the research is to design an adaptive query interface for mining crime data or similar kind of problems. The proposed query interface provides a tool for making an online query and helps in identifying crime hot spots, predict crime trends for the crime hot spots based on the query.

Criminal web data provide unknown and valuable information for Law enforcement agencies continuously. The analysis of vast capacities of comprehensive criminal web data is very complicated in an area over periods of time and that is one of the most significant tasks for law enforcement. Crime database consists of various relational tables which contains the information about crime details in a region under various crime heads such as murder, rape etc. at different



time points. Advanced analytical methods are required to extract useful information from large amount of crime data. Data mining is looked as a solution to such problems [2].

Many scientific researchers have been done on the importance of crime data mining and their results are revealed in the new software applications to analysis and detecting the crime data.

A framework has been developed by Hosseinkhani, et al. [2] for crime web mining consists of two parts. In the first part, some pages which are concerned with the targeted crime are fetched. In the second part, the content of pages are parsed and mined. In fact, a crawler fetches some pages which are associated with the crimes. Previously, pages were fetched by crawler at a time, which was inefficient since the resource was wasted. The proposed model intends to promote efficiency by taking advantage of multiple processes, threads, and asynchronous access to resources.

According to research by Hosseinkhani et al. [5] the aim was suggesting a framework by using concurrent crawler to show the process of exploring the criminal accused of legal data evaluation which insures the reliability gap.

II. WEB CRIME MINING All intelligence-gathering and law-enforcement organizations major challenge is facing to the

efficient and correct evaluating of the crime data growing volumes. One of the examples of this can be complex conspiracies that are often hard to undo since the knowledge of suspects can be geographically span and diffuse in the long time. Detecting cybercrime can be very hard as well, because of frequent online transactions and busy network traffic which create huge amounts of data and just a portion of which relates to illegal activities [2].

In the last decade, through the rapid growth of the Web and through the many unique characteristics, in the following some of them are shown that causes of attracting and challenging for mining the useful information and knowledge [9].

1. Facing to the huge amount of information on the Web that is very wide and diverse so any user can find information on almost anything on the Web.

2. Huge amount of data from all types are exist in unstructured texts, semi-structured Web pages structured tables, and multimedia files.

3. The diversity of the information on the. Multiple pages show similar information in different words or formats based on the diverse authorship of Web pages that make the integration of information from multiple pages as a challenging problem.

4. An association is exist on the significant amount of information of the Web. Hyperlinks are in Web pages across different sites and within a site. Hyperlinks are implicit conveyance of authority to the target pages in across different sites. And hyperlinks serve as information organization mechanisms within a site.



www.elvedit.com 36

5. The information on the Web is noisy that is comes from two main sources. The first one is that a typical Web page involves many pieces of information for instance the navigation links, main content of the page, copyright notices, advertisements, and privacy policies. Only part of the information is useful for a particular application but the rest is considered noise. For performing a fine-grain, the data mining and Web information analysis, the noise should be removed. The second one is due to the fact that the Web does not have quality control of information, for example, a large amount of information on the Web is of low quality because any one can write everything.

6. The Web is about services for example most commercial Web sites allow the users to perform useful operations at their sites such as paying bills, purchasing products, and filling the forms.

7. The Web pages are dynamic that is the information is changes constantly. Copping the changes and monitoring them is an important issue for many applications.

8. The Web is a virtual society that is not only information, data and services; it also is the organizations, the interactions of people, and automated systems. Any user can communicate with people anywhere in the world easily and express his/her views on anything in Internet blogs, forums and review sites [10].

All these characteristics present both challenges and opportunities for mining and discovery of information and knowledge from the Web. This research only focuses on mining textual data. For mining of images, videos and audios please refer to Djeraba, et al. [11], Perner [12]. To explore information mining on the Web, it is necessary to know data mining, which has been applied in many Web mining tasks. However, Web mining is not entirely an application of data mining. Due to the richness and diversity of information and other Web specific characteristics discussed above, Web mining has developed many of its own algorithms [10].

The Web mining process and the data mining process are very similar to each other and their difference is just in their data collection. In traditional method of data mining, the data is gathered and stored in a data warehouse and the other hand, in Web mining, the data gathered is a substantial task that includes crawl of the large number of target Web pages [13]

Web pages are also quite different from conventional text documents used in traditional IR systems. First, Web pages have hyperlinks and anchor texts, which do not exist in traditional documents (except citations in research publications). Hyperlinks are extremely important for search and play a central role in search ranking algorithms. Anchor texts associated with hyperlinks too are crucial because a piece of anchor text is often a more accurate description of the page that its hyperlink points to. Second, Web pages are semi-structured. A Web page is not simply a few paragraphs of text like in a traditional document. A Web page has different fields, e.g., title, metadata, body, etc. The information contained in certain fields (e.g., the title field) is more important than in others. Furthermore, the content in a page is typically organized and



presented in several structured blocks (of rectangular shapes). Some blocks are important and some are not (e.g., advertisements, privacy policy, copyright notices, etc). Effectively detecting the main content block(s) of a Web page is useful to Web search because terms appearing in such blocks are more important [3].

A criminal might either give a deceptive identity or falsely use an innocent person’s identity. There are currently two ways law enforcement officers can determine false identities. First, police officers can sometimes detect a deceptive identity during interrogation and investigation by repeated and detailed questioning, such as asking a suspect the same question (“What is your Social Security number?”) over and over again. The suspect might forget his or her false answer and eventually reply differently. Detailed questioning may be effective in detecting lies, such as when a suspect forgets detailed information about the person whose identity he or she is impersonating. However, lies are difficult to detect if the suspect is a good liar. Consequently, there are still many deceptive records existing in law enforcement data. Sometimes a police officer must interrogate an innocent person whose identity was stolen, until the person’s innocence is proven [3]. Second, crime analysts can detect some deceptive identities through crime analysis techniques, of which link analysis is often used to construct criminal networks from database records or textual documents. Besides focusing on criminal identity information, link analysis also examines associations among criminals, organizations, and vehicles, among others. However, in real life crime analysis usually is a time consuming investigative activity involving great amounts of manual information processing [3].

III. CRIME DATA MINING TECHNIQUES The traditional data mining techniques just classify the patterns in structured data for

example, classification and prediction, association analysis, outlier analysis and cluster analysis. On the other hand, the newer techniques identify patterns from unstructured and structured data [14]. Crime data mining increases the privacy concerns like the other forms of data mining [15]. However, the researchers’ effort to promote the various automated data mining techniques for national security applications and local law enforcement. Particular patterns are identifies by Entity extraction from data such as images, text, or audio materials that has been utilized to automatically identify addresses, persons, vehicles, and personal characteristics from police narrative reports [16]. In computer forensics, the extraction of software metrics [17] which includes the data structure, program flow, organization and quantity of comments, and use of variable name scan facilitate further investigation by, for example, grouping similar programs written by hackers and tracing their behavior. Entity extraction provides basic information for crime analysis, but its performance depends greatly on the availability of extensive amounts of clean input data.

The main techniques of the crime data mining are clustering [18], association rule mining [19], classification [20] and sequential pattern mining [21]. Although all of these efforts, the crime Web mining still is a highly complex task.



www.elvedit.com 38

1. Clustering techniques group data objects into classes by similar characteristics to minimize or maximize interclass similarity for instance, to identify suspects that bearing the crimes in similar ways or discriminate among groups belonging to different gangs. These techniques do not have a set of predefined classes for assigning items. Some researchers use the statistics-based concept space algorithm to automatically associate different objects such as persons, organizations, and vehicles in crime records [22]. Using link analysis techniques to identify similar transactions, the Financial Crimes Enforcement Network AI System [23] [23] exploits Bank Secrecy Act data to support the detection and analysis of money laundering and other financial crimes. Clustering crime incidents can automate a major part of crime analysis but is limited by the high computational intensity typically required.

2. Association rule mining determines frequently occurring item sets in a database and offerings some patterns as rules that been used in network intrusion detection to develop the connection rules from users’ interaction history. Investigators also can apply this technique to network intruders’ profiles to help detect potential future network attacks [24]. Similar to association rule mining, sequential pattern mining finds frequently occurring sequences of items over a set of transactions that occurred at different times. In network intrusion detection, this approach can identify intrusion patterns among time-stamped data. Showing hidden patterns benefits crime analysis, but to obtain meaningful results requires rich and highly structured data.

3. Deviation detection utilizes the particular measures to study data that differs noticeably from the rest of the data. Also called outlier detection, investigators can apply this technique to fraud detection, network intrusion detection, and other crime analyses. However, such activities can sometimes appear to be normal, making it difficult to identify outliers.

4. Classification finds mutual properties between various crime entities and arranges them into predefined classes that have been applied for identifying the source of e-mail spamming according to the sender’s structural features and linguistic patterns [25]. Often used to predict crime trends, classification can reduce the time required to identify crime entities. However, the technique requires a predefined classification scheme. Classification also requires reasonably complete training and testing data because a high degree of missing data would limit prediction accuracy.

5. String comparator techniques that show the relation the textual fields in pairs of database records and calculate the correspondence among the records that can detect deceptive information in criminal records for instance the name and address [26]. the researchers can utilize string comparators to evaluate textual data that often need intensive computation. String comparison is the interesting field for computer scientists that whether string matching or string distance measures. Levenshtein define a usual measure



of similarity between two strings as “edit distance” [27] so, the minimum number of, deletions, single character insertions, and substitutions need to transform one string into the other.

A description of the nodes role in a conceptual network is Social network analysis. Investigators can use this technique to construct a network that illustrates criminals’ roles, the flow of tangible and intangible goods and information, and associations among these entities. Further analysis can reveal critical roles and subgroups and vulnerabilities inside the network. This approach enables visualization of criminal networks, but investigators still might not be able to discover the network’s true leaders if they keep a low profile.

IV. CONCLUSION A web page involving a crime can be thought of as a chain of actions with a series of background attributes. Thus, we can analyze web information from the perspective of events and apply some research results related to events to solve the problem of web crime mining. An event is identified by event triggers, is associated with participants, time, location, et al., and is a larger semantic unit compared with a concept. There is an intrinsic link between events. It is a new attempt to apply the semantic analysis technology of events to mine web crime information on the web. The majority of digital evidence is collected from textual data such as blogs, as e-mails, web pages, text documents and chat logs. The researcher uses some search tools to explore and extract the useful information from the text because the nature of textual data is unstructured and then for further investigation, enter the appropriate pieces into a well-structured database manually which will be boring and error prone. Therefore, the investigators expertise and experience is very important in search and the quality of an analysis. If a criminal hide some essential information, it may be missed. In this review all preliminary concepts such as Web Mining, Criminal Identities and Crime Data Mining Techniques are described. The vision of the Web Mining is to provide a Web where all published material is understandable by software agents. Moreover, Data Mining defined as the process of discovering useful patterns or knowledge from data sources, e.g., databases, texts, images, the Web, etc. Web mining aims to discover useful information or knowledge from the Web hyperlink structure, page content, and usage data. Inspection of files involves searching content for information that can be used as evidence or that can lead to other sources of information that may assist the investigation process and analysis of the retrieved information. It is typically up to the investigator on what and how to search for evidence, depending on the case.

Therefore, we evaluated State-of-the-Art approaches for extracting useful information by means of Data Mining, in order to find crime hot spots out and predict crime trends for them using crime data mining techniques.



www.elvedit.com 40

REFERENCES [1] Fayyad, U.M., and Uthurusamy, R. ( Aug. 2002). Evolving Data Mining into Solutions for Insights.

Comm. ACM. 28-31.

[2] Hosseinkhani, J., Chuprat, S., and Taherdoost, H., (2012a). Criminal Network Mining by Web Structure and Content Mining. 11th WSEAS International Conference on Information Security and Privacy (ISP '12), Prague, Czech Republic September 24-26.

[3] Hosseinkhani, J., Chuprat, S., Taherdoost, H., and Shahraki Moghaddam, Amin., (2012b). Propose a Framework for Criminal Mining by Web Structure and Content Mining. International Journal of Advanced Computer Science and Information Technology (IJACSIT), Helvetic Editions. 1(1). 1-13.

[4] Keyvanpour, M., Javideh, M., and Ebrahimi, M. (2011). Detecting and investigating crime by means of data mining: a general crime matching framework, Elsevier , Procedia Computer Science 3 (2011) 872–880.

[5] Hosseinkhani. J, Taherdoost. H, and Chuprat. S. (2013). Discovering Criminal Networks by Web Structure Mining. 7th International Conference on Computing and Convergence Technology. 3-5 December, Seoul, Korea (South).

[6] Kumar, M., Gupta, A., Saha, S. (2009). Approach to Adaptive User Interfaces using Interactive Media Systems. In: Proceedings of the 11th international conference on Intelligent user interfaces.

[7] Newsome, M., Pancake, C., Hanus, J. (1997). HyperSQL: web-based query interfaces for biological databases. In: Proceedings of the Thirtieth Hawaii International Conference on System Sciences.

[8] Che, D., Aberer, K., Chen, Y. (1999). The design of query interfaces to the GPCRDB biological database. In: Proceedings of User Interfaces to Data Intensive Systems.

[9] Baldi, P., Frasconi, P., and Smyth. P. (2008). Modeling the Internet and the Web: Probabilistic Methods and Algorithms. Wiley.

[10] Liu, Bing. (2007). Web data mining: exploring hyperlinks, contents, and usage data. Springer Verlag.

[11] Djeraba, C. O., Zaiane, R., and Simoff, S.( 2007). (eds.). Mining Multimedia and Complex Data. Springer.

[12] Perner, P. (2003). Data Mining on Multimedia Data, Springer.

[13] Duda, R. O., Hart, P. E. and Stork, D. G, (2001). Pattern Classification. John Wiley & Sons Inc., 2nd edition.

[14] Han, J. and Kamber, M. (2010). Data Mining: Concepts and Techniques, Morgan Kaufmann.

[15] Kargupta, H., Liu, K., and Ryan, J. (2003). “Privacy-Sensitive Distributed Data Mining from Multi-Party Data,” Proc. 1st NSF/NIJ Symp. Intelligence and Security Informatics, LNCS 2665, Springer-Verlag, pp. 336-342.

[16] Chau, M., Xu, J.J., and Chen, H. (2007). “Extracting Meaningful Entities from Police Narrative Reports”, Proc. Nat’l Conf. Digital Government Research, Digital Government Research Center, pp. 271-275.



[17] Gray, A., Sallis, P., & MacDonell, S. (1997). “Software Forensics: Extending Authorship Analysis Techniques to Computer Programs,” Proc. 3rd Biannual Conf. Int’l Assoc. Forensic Linguistics, Int’l Assoc. Forensic Linguistics, pp. 1-8.

[18] Jain, A.K., Murty, M.N., Flynn, P.J. (1999): Data clustering: a review. ACM Computing Surveys 31(3), 264–323.

[19] Agrawal, R., Imielinski, T., Swami, A.N (1993): Mining association rules between sets of items in large databases. In: Proceedings of the ACM SIGMOD International Conference on Management of Data.

[20] Han, J., Kamber, M.(2009): Data mining: concepts and techniques. Morgan Kaufmann, San Francisco.

[21] Agrawal, R., Srikant, R. (1995): Mining sequentiel motifs. In: 11th Int’l Conf. on Data Engineering.

[22] Hauck, R.V. et al., (2002). “Using Coplink to Analyze Criminal-Justice Data” Computer, Mar, pp. 30-37.

[23] Senator, T. et al., (1995). “The FinCEN Artificial Intelligence System: Identifying Potential Money Laundering from Reports of Large Cash Transactions”, AI Magazine, vol.16, no. 4, pp. 21-39.

[24] Lee, W., Stolfo, S.J., and Mok, W. (1999). “A Data Mining Framework for Building Intrusion Detection Models,” Proc. 1999 IEEE Symp. Security and Privacy, IEEE CS Press, pp. 120-132.

[25] Vel, O. de. et al., (2001). “Mining E-Mail Content for Author Identification Forensics,” SIGMOD Record, vol. 30, no. 4, pp. 55-64.

[26] Wang, G., Chen, H., and Atabakhsh, H. (2007). “Automatically Detecting Deceptive Criminal Identities,” Comm. ACM, Mar., pp. 70-76.

[27] Levenshtein, V.L. (1966). Binary codes capable of correcting deletions, insertions, and reversals. Soviet Physics Doklady 10, 707–710.

[28] Hosseinkhani, J., Ibrahim, S., Chuprat, S. and Hosseinkhani, N.J. Web Crime Mining by Means of Data Mining Techniques, Research Journal of Applied Sciences, Engineering & Technology (Print ISSN: 2040-7459 Online ISSN: 2040-7467), 2013.


Develop a New Method for People Identification Using Hand Appearance

Authors

Mahdi Nasrollah Barati Department of Computer Engineering, Islamic Azad University, Babol Branch, Iran

Seyed Yaser Bozorgi Rad Department of Computer Engineering, Islamic Azad University, Babol Branch, Iran

[email protected] Babol, Iran

[email protected], Iran

Abstract

In this paper a new method for people identification using hand appearance is presented. In this method, the contour information is used for matching. For this purpose, after applying pre-processing algorithms and edge detection, contour extraction, and to help the offices of concentric, hand’s information including the number of pixels is limited to offices, will be extracted. By using extracted information, Matching will be done in the database identity and person will be identified. Benefits of the proposed method can be its lack of sensitivity to rotating and zooming the image pointed out. Practical results will show the accuracy of this method for identification. The proposed method can be used in other fields such as curve matching in addition to hand geometry identification.

Key Words

Identification, contour extraction, Matching, Hand appearance.

I. INTRODUCTION

Nowadays identifying the people is the most important world security issues. Different methods are using for identification, such as passport, but there are some problems on using it. Also identifying using computer technology is the main objectives of governments. By these reason, scientists are searching for features in the human body that could be use to identify individuals.

Develop a New Method for People Identification Using Hand Appearance Mahdi Nasrollah Barati and Seyed Yaser Bozorgi Rad


Examples of these features include iris [1], finger [2], face appearance [3], signature [4] and the appearance. In fact, the features of body can help in this way.

One feature of the human body that can identify its utilization is a hand’s human appearance. A variety of methods available in matching appearance are presented. Some of these methods in matching use hand geometry features such as fingers and fingertips [5]. In these Methods, fingertips and toes and also the junction of the hand fingers have been used as points of review in determining the characteristics. There are also other methods that use existing lines and patterns on hand to compare and extract the features required [6]. These lines include lines of finger and palm lines. In addition, some methods use combination of these two methods. In such ways the geometric characteristics of hands and lines in the palms and fingers are used.

Innovative approach in this article are provided with matching curves obtained from the edge detection images using drawing concentric series of offices and counting the number of points in between these offices as a characteristic image to match its looks. Please note that the former method to compare with this method was implemented [7] and doing some tests. It was found that this algorithm has better and more acceptable results that in practical results section I will show it. In addition, this method is not sensitive to zoom and rotation.

II. PROVIDING IMAGES DATABASE

To collect images of people, two methods have been used. First with the help of a scanner, hands placed on the scanner and it scans. In this way the background will be black, like Figure 1. In the second method, images are used from a digital camera. At first, the hand is placed in front of a white paper and then photos will be taken. In this way, the background image is white. Another important issue related to data collection should be noted is that in all images the distances between each finger must be equal. The fingers distance will set by eye in this paper.

Images obtained by cameras or scanning devices for edge detection are used in the preprocessing stage. Edge detection algorithm can be algorithms such as sable edge detection method [8].

Edge detected Image may have extra edges. For removing these extra edges, greater threshold amount in edge detection is used. However, it should be noted that the high threshold might not be remove extra edges completely. In edge detected image that is shown in figure 2, the thinning methods have been used.



www.elvedit.com 44

FIGURE 1: SCANNED HAND

FIGURE 2: EDGED IMAGE

Thinning results can be seen in Figure 4. In thinning phase, morphological methods have been used based on the review of each pixel's neighbors without rupturing in the thin curve [9]. In this method, a considered for each pixel of the edge is concerned. Then this pixel’s connectivity coefficient is calculated. If the connectivity coefficient obtained is equal to 1 and isn’t end point, it will be removed. Otherwise, without any changes could go to next pixel. This work should be done for all black pixels in the image.

As you can see in Figure 3, there is no tear in the picture. One advantage of this method is that the end points of curve can be detected and removing end points can prevent.



FIGURE 3: THINNED IMAGE

III. EXTRACTING HANDS FEATURES Here, the hands features are extracted. These features should have enough information, In

addition they should be stable in image rotation, zooming and moving. To eliminate rotation and zooming image, the suitable axis in the hand image is required that images rotation and zooming could be tolerated. For finding this axis, two appropriate points in hands image are selected. The first point that has been used for this purpose is the center of edges mass that its position is Unchanged in zooming and rotation. The purpose of the mass center is the point that X and Y coordinates is equal to average total black pixel in image. The center of mass can be calculated through the average of X and Y components of black pixels in the image with the help of equation (1).

N

xxcenter i (1)

N

yycenter i

Here, another point for the axis is needed. Farthest point to the center of mass as the second axis point is chosen. Whereas point choosing is very important in the matching algorithm, all probabilities are considered for the farthest point. In other words, many points choose as possible for the farthest point. Selection algorithm for the farthest possible points is as follows:

Farthest point on the contour to the center of mass is calculated named the distance between that and the center of mass as‘d’. All points that the distance to the center is larger than 0.9d kept and the rest points are removed.

The continuous component extraction algorithm is applied to the remaining parts and continuous components have been extracted. In each continuous component the point that has the greatest distance is selected to the center of mass as possible point.



www.elvedit.com 46

After finding the farthest possible points, for each of these points, the features extracted and stored. Now features should be extracted for every possible point. For this, concentricity circles is drawn that the center is the possible point. Radius should be increased by a fixed step as u see in figure (4). Choosing distance between the radiuses is particularly important. This algorithm will be more accurate if the small distance is selected, but it makes the algorithm slower. In addition, decreasing the distance make the algorithm less Accurate. So selecting a good distance is one the important part of this algorithm. The appropriate distance experimentally should be obtained.

FIGURE 4: CONCENTRICITY CIRCLES

After drawing offices, black pixels are counted that are between the circles and store it in a one-dimensional array as hand’s features. However, due to the several possible points, there are several features array. So, these arrays could be used for hands matching.

Features that are extracted in the previous step are stable to the rotation but they are not stable to the zooming. To eliminate this problem, the numbers in the previous step in the d (distance between center of mass to the farthest point) is divided, so the features extracted are stable to zooming.

IV. MATCHING METHOD After preparing the database and extracting features for different people's hands, it can be used

to identify individuals. For this purpose the matching algorithm is used. So, the input image data is compared to the information’s that stored in the database and if they matches the person’s identity information is extracted. Different stages of the matching algorithm are as follows:

Edge of Input image and the contour of hand should be extracted. Now all necessary steps such as thinning and waste edge removal can be applied.

The features of the hand can extract by using mass center and the farthest points.



Similarity coefficient of input image characteristics and features in the database is calculated. A variety of methods exist to obtain the similarity coefficient that in this paper, SSD method is used for calculating the similarity coefficient. This method is formulated and is shown by Eq.(2).

n

ifeatureifeaturessd

])[1][2(

(2)

The Number ‘n’ is the number of features that has been extracted. This division is for normalizing the number obtained to the number of features. Also feature1 [i] is the ith feature of first image and feature2 [i] is ith feature of second image. Note that all probability for farthest point is considered, for each image several feature groups is extracted. So all possible modes and the minimum ‘ssd’ is the similarity measure of two images is concerned.

Using the coefficients obtained in the previous step, an image in database with the highest similarity (lowest ssd) with the input image should be selected. If ssd rate calculated for the selected image is less than a threshold, Input image with the database image are considered similar.

V. NUMERICAL RESULT For The results of this algorithm 120 samples from different people's hand images in

experiments have been used. The implementation of this algorithm is in Visual studio 6 environment and by C++ language. PC which has been used in the testing is Intel ® Pentium ® 4-2.6GHz-256MB of RAM. To show lack of sensitivity of the algorithm to zooming and rotating, the zoomed and rotated images of different people in experiments have been used.

The value that radius should be increase can be obtained experimentally. This value equal to 30 pixels in the tests carried out. The size of images is 600 * 600 pixels and the file format is Bitmap. The proposed algorithm by two methods has been tested. In the first method, input image in the images collection was searched. In the second method, input image in collection which the input image isn’t on it was searched. In the fist way the objective is the understanding that if the method can know that this image is not in collection. Considering the selected parameters, the success rate is 100 percent. This means all images matched right. While in previous methods the success rate is 94 percent [7]. Percentage of success in the second way is equal to 97 percent. It should be noted however, many existing methods are sensitive to zooming and rotation.



www.elvedit.com 48

FIGURE 6: SUCCESS RATE OF PREVIOUS METHODS

VI. CONCLUSION In this paper a new method for identification using hand geometry is presented. In this

method, the contour information is used for matching. To this end, by applying a pre-processing algorithms and edge detection contour of the hands image extracted and by circles obtained the feature of the hands. The features are the number of pixels between these circles. Extracted features should be matched to collected hands image and the inputs hands image could be identified. Benefits of the proposed method can be its lack of sensitivity to rotate and zoom the image. Due to experimental results and precision of this method, this method can be used for identification.

Suggestions in improving this algorithm are also provided. For example, other lines in the palm and fingers in matching could be used. Therefore, the result of this method could be improved by using other features even in the larger collections of images. In addition, a variety of applications for these methods is available. This method could be used in matching of signatures, geographic maps, ear and lip.

REFERENCES

[1] W.W. Boles and B. Boashah, “A Human Identification Technique Using Images of the Iris and Wavelet Transform”, IEEE Trans. on Signal Processing, Vol.46, pp.1185-1188, 1998.

[2] A.Jain, Y.Chen and M.Demirkus, "Pores and Ridges: Fingerprint Matching Using Level 3 Features", Proc. Of IEEE ICPR conference, pp. 477-480, 2006.

[3] A.M. Martínez, "Recognizing Imprecisely Localized, Partially Occluded, and Expression Variant Faces from a Single Sample per Class", IEEE Transactions on Pattern Analysis and Machine Intelligence, vol.24 no.6, pp.748-763, June 2002 .



[4] M.Kondo and D.Muramatsu and M.Sasaki and T.Matsumoto , "Bayesian MCMC for Biometric Person Authentication Incorporating On-line Signature Trajectories", International Conference in Biomechanics , 2007.

[5] L.N.Wong and P.Shi , "Peg-Free Hand Geometry Recognition Using Hierarchical Geometry and Shape Matching", online: www.ee.ust.hk/~eeship/Papers/MVA02.pdf

[6] X.Wang, H.Gong, H.Zhang, B.Li and Z.Zhuang,"Palmprint Identification using Boosting Local Binary Pattern", Proc. Of IEEE ICPR conference, pp.503-506, 2006.

[7] M.A.Khaiyat and F.Kamangar, “Planar Curve Representation and Matching,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 76019, May 1990.

[8] R.C.Gonzalez and R.E.Woods, Digital Image Processing Prentice Hall, 2nd edition, 2002.

[9] J.R.Parker, Algorithms for Image Processing and Computer Vision. Wiley Pap/Cdr edition 1996.

International Journal of Advanced Computer Science and Information Technology (IJACSIT) Vol. 3, No.1, 2014, Page: 50-69, ISSN: 2296-1739 © Helvetic Editions LTD, Switzerland www.elvedit.com

Model and Solve the Bi-Criteria Multi Source Flexible Multistage Logistics Network

Authors

Seyed Yaser Bozorgi Rad Department of Computer Engineering, Islamic Azad University, Babol Branch, Iran


Mohammad Ishak Desa Department of Modeling and Industrial Computing, Faculty of Computer Science & Information System. Universiti Teknologi Malaysia (UTM)

[email protected] Johor, Malaysia

Sara Delfan Azari Ayandegan Institute of Higher Education, Tonekabon, Iran

[email protected] Tonekabon, Iran

Abstract

Flexible Multistage Logistics Network (fMLN) is an extension of the traditional multistage logistics network whereby a customer can procure goods directly from plants or distribution centers needless of retailers. This research intends to formulate the bi-criteria multi source single product fMLN model and discover methods to solve it. Here, total logistics cost and total product delivery time should be minimized simultaneously. By far, fMLN problems have been dealt with in single source form, meaning each customer could only be served by only one source. Because this issue is NP-hard, meta-heuristic techniques such as Genetic Algorithm (GA) have been used to solve the problem. However, under realistic settings, fMNL is multi-source, meaning each customer may be served by a number of facilities simultaneously. Because a multi-source fMNL problem is more complex than the single source in terms of both options as well as constraints, GA will also require enhancement. The proposed solution of this research is representing the chromosome in a new state, capable of improvising the constraints of the problem by a considerable ratio and with the defined crossover and mutation to solve the general bi-criteria multi-source fMNL. The obtained result using enhanced GA will show that it is dramatically improved comparing with using standard GA in order to having lower cost and time.

Key Words

Bi-criteria multi source Flexible Multistage Logistics Network (fMLN), Genetic Algorithms, Multi–objective optimization, PARETO solution.

Model and Solve the Bi-Criteria Multi Source Flexible Multistage Logistics Network Seyed Yaser Bozorgi Rad, Mohammad Ishak Desa, and Sara Delfan Azari


I. INTRODUCTION Gen et al. [11] claimed that although the traditional multistage logistics network model and its

application had made a big success in theory and business practices, the traditional structure of logistics network is unable to fit very well with the fast changing competitive environments and meet the diversified customer demands. Therefore, Gen et al. [11] introduced three new delivery modes in which the goods can move from plants to retailer directly not via distribution centers, or sometimes the customer can get the goods from plant or from distribution center directly not via retailer. The authors called this new logistics network as the flexible Multistage Logistics Network (fMLN) as it is shown by Figure 1.

FIGURE 1: THE STRUCTURE OF FLEXIBLE MULTISTAGE LOGISTICS NETWORK (FMLN) MODELS [11].

Gen et al. [11] indicated that the bi-criteria linear logistics model (or bi-criteria transportation problem: bTP) is a special case of multi-objective logistics model since the feasible region can be depicted with a two dimensional criteria space. The following two objectives are considered:



www.elvedit.com 52

minimizing the total logistics cost and minimizing the total delivery time.

Rajabalipour et al, [16] considered two- stage logistic networks comprised of potential suppliers, distributing centers (DCs) and also actual consumers at the first level. Each consumer has pre-specified demand of single item product for a period of time (e.g. season, year and, etc.) and the network could be flexible with potential (probably expensive) direct shipments only from the supplier to the consumers.

Recently, GAs has been successfully applied to logistics network models. Michalewicz et al. [15] and Viagnaux and Michalewicz [19] are among the first who discussed the use of GA for solving linear and nonlinear transportation problems. In their study, while matrix representation was used to construct a chromosome, the matrix-based crossover and mutation had been developed. Another (GA) approach for solving solid TP was given by Li et al. [13]. They used the three dimensional matrix to represent the candidate solution to the problem. Syarif and Gen et al. [18] considered production/distribution problem modeled using tsTP and proposed a hybrid genetic algorithm. Gen et al. [10] developed a priority-based Genetic Algorithm (priGA) with new decoding and encoding procedures considering the characteristic of tsTP. Altiparmak et al. [3] extended priGA to solve a single-product multi-stage logistics design problem. The objectives are minimization of the total cost of supply chain, maximization of customer services that can be rendered to customers in terms of acceptable delivery time (coverage), and maximization of capacity utilization balance for DCs (i.e. equity on utilization ratios). Furthermore, Lin et al. [14] proposed a hybrid genetic algorithm to solve the location-allocation model’s problem of logistic network, and Altiparmak et al. [2] also apply the priGA to solve a single-source, multi-product multi-stage logistics design problem. As an extended multi-stage logistics network model, Lee et al. [12] apply the priGA to solve a multi-stage reverse logistics network problem (mrLNP), minimizing the total costs to reverse logistics shipping cost and fixed cost of opening the disassembly centers and processing centers. Gen and Syarif [18] proposed a new approach called spanning tree-based hybrid genetic algorithm (hst- GA) to solve the multi-time period production/distribution and inventory problem (mt-PDI). Costa et al. [5] presented an innovative encoding–decoding procedure embedded within a genetic algorithm (GA) to minimize the total logistic cost resulting from the transportation of goods and the location and opening of the facilities in a single product three-stage supply chain network.

For any optimization problem, there is an optimization criterion (i.e. evaluation function) to be minimized or maximized. The evaluation function represents a measure of the quality of the developed solution. Searching the space of all possible solution is a challenging task. An additional constraint on the domain of search for the parameters makes the problem quite difficult. The constraints might affect the performance of the evolutionary process since some of the produced solutions (i.e individuals) may be unfeasible. Unfeasible solution represents a waste of computation effort. In fact, it was reported that no general methodology to handle constraints exist although several methods were introduced. Rejecting unfeasible individuals, penalizing unfeasible individuals or moving these individuals to the feasible domain are among the many



methods proposed [17].

There are some approaches to handle the constraints optimization problems such as death penalty, static penalties, dynamic penalties, GENOCOP system, Behavioral memory and etceteras [21]. For some similar problems to fMLN with high constraints some researchers tried to add some heuristic rules to GA to satisfy the problem constraints and obtain a good solution. Yaohua and Chi [20] proposed a random search based on heuristic rules and a dynamic rule selection method based on GA to solve large size single-stage batch scheduling problem and Alim and Ivanov [1] proposed some heuristic rules embedded GA to solve In-Core Fuel Management Optimization Problem. Craenen et al. [6] compared three different heuristics based Evolutionary Algorithm (EA) on the same problems and suggested the best one to solve the constraints optimization problems.

The general objectives of this paper are to formulate the bi-criteria multi-source single product flexible Multistage Logistics Network (fMLN) problem and to discuss the algorithms that we have developed to solve it.

II. MATERIALS AND METHODS Bi-Criteria Multi Source Single Product fMLN Model

In general, a bi-criteria multi source flexible multistage logistics network (fMLN) problem is to establish the optimum product amount shipped from plants to the customers and the best product delivery routes to fulfill the customer’s order with the optimum product delivery time in all network phase that reduce the total logistics network costs. The mathematical model of the bi-criteria (fMLN) is developed with the following assumptions:

1. Single product and single time period (week, month, season, year or etc) case of a logistics network optimization problem is considered.

2. There are a maximum of four levels: plants, DCs, retailers and customers.

3. There are three delivery modes: normal delivery, direct shipment and direct delivery

4. Every customer, retailer and distribution center (DC) can be served by multi facilities. There is no preference for retailers, DCs and customers to provide orders, consequently fulfilling orders through multi sources at once.

5. Customer demands are known in advance.

6. Customers will get products at the same price, no matter where he/she gets them; it means that the customers have no special preferences.



www.elvedit.com 54

The following notations are used to formulate the model:

Notation:

Indices:

i: index of plant (푖 = 1,2,3,⋯ , 퐼) j: index of DC (푗 = 1,2,3,⋯ , 퐽) k: index of retailer (푘 = 1,2,3,⋯ ,퐾) l: index of customer (푙 = 1,2,3,⋯ , 퐿)

Parameters:

I number of plants J number of DCs K number of retailers L number of customers P Plant i DC DC R Retailer k C Customer l 퐵 Output of plant i d Demand of customer l C Unit shipping cost of product from P to DC C Unit shipping cost of product from DC to R C Unit shipping cost of product from R to C C Unit shipping cost of product from P to 퐶 C Unit shipping cost of product from DC to 퐶 C Unit shipping cost of product from P to R T Shipping time per lot of product from P to 퐷퐶 T Shipping time per lot of product from 퐷퐶 to 푅 T Shipping time per lot of product from 푅 to 퐶 T Shipping time per lot of product from 푃 to 퐶 T Shipping time per lot of product from 퐷퐶 to 퐶 T Shipping time per lot of product from 푃 to 푅 u Upper bound of the capacity of DC u Upper bound of the capacity of R f Fixed part of the open cost of DC C Variant part of the open cost (lease cost) of DC q Throughout of DC 푞 = ∑ 푋 ,∀푗 f Open cost of DC 푓 = 푓 + C 푞 ,∀j g Fixed part of the open cost of R C Variant part of the open cost (lease cost) of R



q Throughout of R Piet S 푞 = ∑ 푋 ,∀푘 Pie Slats Piet A. Slats g Open cost of R 푔 = g + C 푞 ,∀k

Decision Variables:

푋 Product amount shipped from P to DC 푋 Product amount shipped from DC to R 푋 Product amount shipped from R to C 푋 Product amount shipped from P to 퐶 푋 Product amount shipped from DC to 퐶 푋 Product amount shipped from P to R

푦 = 1,푖푓퐷퐶 푖푠표푝푒푛0, 표푡ℎ푒푟푤푖푠푒

푦 = 1,푖푓푅 푖푠표푝푒푛0,표푡ℎ푒푟푤푖푠푒

푊 = 1, 푖푓푋 > 00,표푡ℎ푒푟푤푖푠푒






The two objective functions are to minimize the total logistic cost 푍 and total product delivery

time 푍 . The mathematical model for the fMLN problem is given as follows:

푀푖푛푍 = ∑ ∑ 퐶 푋 + ∑ ∑ 퐶 푋 + ∑ ∑ 퐶 푋 + ∑ ∑ 퐶 푋 + (1) 2݇ݕ݇݃ܭ1=݇ +1݆ݕ݆݂ܬ1=݆ +6݅݇ܺ6݅݇ܥܭ1=݇ܫ1=݅ +5݆݈ܺ5݆݈ܥܮ1=݈ܬ1=݆

Min 푍 = ∑ ∑ 푇 푊 +∑ ∑ 푇 푊 + ∑ ∑ 푇 푊

+ ∑ ∑ 푇 푊 + ∑ ∑ 푇 푊 + ∑ ∑ 푇 푊 (2)



www.elvedit.com 56

Subject to:

∑ 푋 +∑ 푋 + ∑ 푋 ≤ 푏 , ∀i (3)

∑ 푋 = ∑ 푋 + ∑ 푋 , ∀j (4)

∑ 푋 + ∑ 푋 = ∑ 푋 , ∀k (5)

∑ 푋 +∑ 푋 + ∑ 푋 ≥ 푑 , ∀푙 (6)

∑ X ≤ u y , ∀j (7)

∑ X ≤ u y , ∀k (8)

푋 ,푋 ,푋 ,푋 ,푋 ,푋 ∈ 푁 , ∀i, j, k, l (9)

where;

푁 = {0, 1, 2, 3,⋯ }

y , y ∈ {0,1},∀j, k (10)

푊 , 푊 , 푊 , 푊 , 푊 , 푊 ∈ {0,1},∀푖, 푗,푘, 푙 (11)

Where there are two objective functions that the objective function Eq. 1 means to minimize the total logistic cost and Eq.2 means to minimize the total product delivery time. The constraint in Eq.3 represents the production limit of plants. The constraints in Eq.4 and Eq.5 are due to the flow of conservation principle. The constraint in Eq.6 ensures that the customers’ demands will be satisfied. The constraints in Eq.7 and Eq.8 ensure that the upper bound of the capacity of DCs and retailers cannot be surpassed. Eq.9 shows that the decision variables related to product amount are non-negative integer.

Proposed Route Based GA (RB-GA) to Solve Bi-Criteria Multi Source Single Product fMLN Problem

A tree-based representation is known to be one way for representing network problems [2]. There are three ways of encoding tree:

(1) Vertex-based encoding

(2) Edge-based encoding

(3) Edge-and-vertex encoding



Gen et al. [11] used vertex-based encoding to solve single product fMLN problem. Using this chromosome representation, if the total demand to the plant exceeds its supply capacity, the customer is assigned to another plant with sufficient products supply and the lowest transportation price between the plant and the customer. According to the above representation, solution is based on finding the best route for delivering the product to each customer when the network is single source in the last layer. Based on above mentioned solution the customer is not allowed to split the order to be fulfilled from different sources simultaneously. The length of every chromosome here is equal to: 3 × 퐿.

It is argued that the solution proposed by Gen et al. [11] is not useful to solve multi source fMLN problem. Here, Edge-and-vertex encoding is used to solve bi-criteria multi source fMLN problem and the new algorithm namely Route Based GA (RB-GA) is developed. Figure 2 represents the chromosome with edge- and vertex encoding. In a normal shipment which the number of plants, DCs, and retailers are I, J and K , representing the number of possible routes for product delivery from plant to each customer, using permutation theory is given by: 퐼 × 퐽 × 퐾.

The total number of possible routes in fMLN = the number of routes for normal delivery + the number of routes for direct shipment + the number of routes for direct delivery = 퐼 × 퐽 × 퐾 + 퐼 × 퐾 + 퐼+ 퐼 × 퐽 = 퐼 × (퐽 + 1) × (퐾 + 1)

Therefore, the total number of possible routes for each customer named 푁푂푅 and is given by:

푁푂푅 = 퐼 × (퐽 + 1) × (퐾 + 1)

FIGURE 2: EDGE-AND-VERTEX ENCODING

Referring to Figure 2, it is obvious that the demand of customer l (푑 ) is distributed to the possible routes which are pertinent to customer l with random amount between 0 and 푑 . Therefore;

푑 = 푑 + 푑 + 푑 + ⋯+ 푑 where 푑 ,푑 ,푑 ,⋯푎푛푑푑 are generating as follows:

푑 = 푟푎푛푑표푚푖푧푒(푑 ) , ∀푙



www.elvedit.com 58

푑 = 푟푎푛푑표푚푖푧푒(푑 − 푑 ) , ∀푙

푑 = 푟푎푛푑표푚푖푧푒(푑 − (푑 + 푑 ) , ∀푙

⋮

푑 = 푟푎푛푑표푚푖푧푒(푑 − ∑ 푑 ) , ∀푙

푑 = 푑 − ∑ 푑 , ∀푙

There are several units of the chromosome for every customer (the number of 푁푂푅 ) which every unit indicates one possible product delivery route to a customer. Here, to simplify and decrease the number of gene especially for large size problem case, every unit is changed to be one gene of the chromosome. Therefore, every gene shows a possible route for customer l with the amount of customer l’s demand as follows:

where

푀 indicates the q th possible route to fulfill the l th customer order, that is, the amount product shipped to customer l through the q th route, where q= 0, 1, 2, ⋯,NOR.

Therefore, Figure 2 can be changed to Figure 3 as below:

FIGURE 3: PROPOSED RB-GA CHROMOSOME REPRESENTATION

푑 is the l th customer demand and 푀 ,푀 ,⋯ ,푀 푎푛푑 푀 could be generated as follows:

푀 = 푟푎푛푑표푚푖푧푒(푑 ) , ∀푙

푷풊 푫푪풋 푹풌 풅풍풒 푴풍

풒



푀 = 푟푎푛푑표푚푖푧푒(푑 − 푀 )

푀 = 푟푎푛푑표푚푖푧푒(푑 − ∑ 푀 ) , ∀푙

푀 = 푑 − ∑ 푀 , ∀푙

therefore

푑 = 푀 + 푀 + 푀 + ⋯+ 푀

In fMLN with J number of DCs and K number of retailers are, the q th route for the l th customer is defined by the following procedures:

i) Let w be the quotient of q ÷ (K+1) and s is the remainder of (q ÷ (K+1)). ii) Let q be the quotient of w÷ (J+1) and r is the remainder of (w ÷ (J+1).

The ID of plant involved q th route is: q+1, which indicates the first node of the route.

The ID of DC involved q th route is: r, which indicates the second node of the route.

The ID of retailer involved q th route is: s, which indicates the third node of the route.

For further illustration suppose that I=2 which is the number of plants, J=2 which is the number of DCs and K=3 which is the number of retailers. Here, the total number of possible routes for each customer is calculated as: 퐼 × (퐽 + 퐼) × (퐾 + 1) = 2 × (2 + 1) × (3 + 1) = 24 . Suppose that the 17 th route is required, therefore, according to above procedures this route is defining as follows:

17 ÷ (3 + 1) w = 4 and s=1

4 ÷ (2 + 1) q = 1 and r =1

Therefore;

The ID of plant involved 17 th route is q+1= 2, the ID of DC involved this route is r =1 and the ID of retailer involved this route is s = 1, that is:

Plant 2 DC 1 Retailer 1 Customer l

Subsequently the following decision variables could be defined as well:



www.elvedit.com 60

푋 which is product amount shipped from Plant 2 to DC 1.

푋 which is product amount shipped from DC 1 to Retailer 1.

푋 which is product amount shipped from Retailer 1 to Customer l

Based on above explanation, all types of decision variables (푋 ,푋 ,푋 ,푋 ,푋 ,푋 ) and their values can be obtained using the following procedures (Figure 4):

Procedure: Define the product amount shipped at every arc of fMLN. Input: Number of Plants (I), number of DCs (J), number of Retailers

(K) and the number of Customer Output: Decision variables and their values for l = 1 to L for i =1 to I for j= 0 to J for k = 0 to K

if j ≠0 we have: 푋 if k≠0 we have: 푋 and 푋 else we have: 푋 end

else if k≠0 we have: 푋 and 푋 else we have: 푋 end

end end end end end Output: Decision variables and their values

FIGURE 4: PSEUDO-CODE TO DERIVE DECISION VARIABLES FROM POSSIBLE ROUTES



It is obvious that to serve every customer at least one plant is needed. Considering the multi source assumption where each customer can be served by multi facilities the ID of plants must be non zero while the ID of DCs and retailers could be zero since for the flexible logistics model, there is direct shipment or direct delivery. Therefore, the customer would be able to split his/her order to be fulfilled from different facilities. Based on the above mentioned chromosome representation (Figure 4), the total number of genes for every customer is equal to:

퐼 × (퐽 + 1) × (퐾 + 1)

(I is the total number of plants, J is the total number of DC and K is the total number of retailer). Subsequently the total number of genes for every chromosome is calculated as follows:

[퐼 × (퐽 + 1) × (퐾 + 1)] × 퐿 (L is the total number of Customers).

Every gene is one set of ID of a plant, ID of DC and ID of retailers with the part of amount of customer demand. The NOR number of gene constitutes one unit for every customer where each unit represents all possible delivery routes to a customer with amount of customer demand for each route from the plant via DC and retailer.

Using this encoding method, an infeasible solution may be generated, which violates the facility capacity constraints, where the penalty method could be useful. As it was mentioned earlier about difficulty for satisfying the two main constraints which are:

∑ 푋 = ∑ 푋 + ∑ 푋 , ∀j

∑ 푋 + ∑ 푋 = ∑ 푋 , ∀k

Proposed RB-GA could satisfy them simply by embedding to the chromosome representation. The above mentioned constraints depict that the summation of incoming product amount to each facility (DC or Retailer) must be equal to summation of out coming product amount from the same facility.

Using proposed RB-GA; define the all possible routes for product delivery to each customer which every route contains the constant amount of the product at all stage of network. It is obvious that every route which passes each DC or retailer has the same amount of incoming and out coming in that DC or retailer. Therefore the summation of all incoming of all routes are equal to the summation of all out coming of all routes at every DC or retailer. In conclusion, it is true that the above equality constraints are satisfied simply.



www.elvedit.com 62

By this chromosome representation the best delivery route/s for each customer and the optimum product amount for every stage of the network will be found. Furthermore the decision about open/close facilities will be made.

Crossover for Edge-and-Vertex Encoding (RB-GA):

Figure 5 shows an example of proposed crossover in this research. It randomly selects two cutting points and then exchanges the substrings between the two parents.

Cutting point 1 = Randomize (L)

Cutting point 2 = Randomize (L)

where; L = total number of customers.

FIGURE 5: AN EXAMPLE OF PROPOSED CROSSOVER



The cross points located at the beginning of every unit, therefore, after crossover, an infeasible solution may not be generated and repairing procedure is not needed.

Mutation for Edge-and-Vertex Encoding (RB-GA)

Here, some offspring will be selected according to mutation rate and the mutation procedure is explained as below:

1- Generate randomize (L-1) +1 // l ih customer will be found// 2- Generate randomize (푑 ) denoted as m_n . 3- Generate randomize (NOR) denoted as m_u, and NOR is the total possible routes for

product delivery to each customer // the number of specific gene of l ih customer that must be mutated will be found //

4- m_u m_n 5- Generate randomly the number of route (randomize (NOR) ) denoted as A, then set g_m =

RT – A // the number of second specific gene of l ih customer that must be mutated will be found //

6- g_m m_n – randomize (m_n)

It is noted that using the proposed crossover and mutation, the chromosomes still would be able to satisfy the equality constraints.

As it was mentioned before, the bi-criteria fMLN problem is a special case of multi objective optimization. In general, with the multiple criteria/objectives, it is impossible to obtain a distinctively optimal solution for all the proposed models. This means that search techniques are required to search a set of concession solutions first, followed by the part where the decision maker uses the preference relation to rank them.

Genetic Algorithms had already been used to solve multi objective problems. Here, RB-GA is developed to solve bi-criteria multi source single product fMLN problem. In principle, multiple objective optimization problems are very diverse from single objective optimization problems. For single objective case, one tries to obtain the best solution, which is totally superior to all other options. In the case of multiple objectives, a solution that is the best with respect to all objectives does not necessarily exist mainly because of the incommensurability and conflict among objectives. A solution may be the best in one objective but the worst in other objectives. Therefore, more often than not exists a set of solutions for the multiple objective cases which cannot simply be compared with one another. Such type of solutions can be named as non-dominated solutions or Pareto optimal solutions, for which no improvement in any objective function is achievable devoid of sacrificing at least one of the other objective functions.



www.elvedit.com 64

III. RESULT As a matter of fact, in bi-criteria fMLN problem there are two objective functions represented

as 푓 and 푓 that must be evaluated in the selection step of GA as following:

푓 = 1 1 1 2 2 3 31 1 1 1 1 1

I J J K K L

ij ij jk jk kl kli j j k k l

Min Z C X C X C X

1 24 4 5 5 6 6

1 1 1 1 1 1 1 1

I L J L I K J K

il il jl jl ik ik j j k ki l j l i k j k

C X C X C X f y g y

푓 = Min 2Z = 1 11 1

I J

ij iji j

T W + 2 2

1 1

J K

jk jkj k

T W + 3 3

1 1

K L

kl klk l

T W + 4 4

1 1

I L

il ili l

T W + 5 5

1 1

J L

jl jlj l

T W + 6 6

1 1

I K

ik iki k

T W

Where 푓 shows the total logistics cost and 푓 shows the total product delivery time of fMLN. In this case we try to obtain a PARETO optimal solution as it is shown in Figure 6.

FIGURE 6: AN EXAMPLE OF PARETO OPTIMAL SOLUTION OF BI-CRITERIA PROBLEM [4].

PARETO optimality characterizes the boundary of solutions that can be achieved by trading-off clashing objectives in the most favorable approach. From here, a decision maker (be it a human or an algorithm) is able to finally select the settings that suit best according to his judgment. The notation of optimal in the Pareto logic is sturdily according to the meaning of supremacy and dominance: An element x1 dominates (is preferred to) an element x2 if x1 is better than x2 in at least one objective function and not worse with reverence to the rest of the objectives. Pareto Optimal definition is an element 푥∗∈ X is Pareto optimal (and hence, part of the optimal set 푋∗)



if it is not dominated by any other element in the problem space X. When it comes to Pareto optimization, 푋∗ is called the Pareto set or the Pareto Frontier.

In the selection and evaluation part RB-GA there are two dissimilar values of fitness function for each chromosome. One is the delivery cost value (푓 ) and the second one is delivery time value (푓 ) of each chromosome. In order to attain PARETO solution it require to obtain the optimal front (front 1) that these solution are not dominated by the other feasible solutions. Figure 6 shows the Pseudo-code of fitness evaluation part of proposed algorithms selection part.

Procedure: Selection of bi-criteria fMLN Input: two objective functions (푓 푎푛푑푓 ) and size of them (m), old Population (N) */Total number of chromosomes ( Population size + all offspring)/* Output: new Population Input: Cost and time of every chromosome N= Population size + all offspring S(1 to N)=N+1; */ defining the fronts in problem space based on fitness function,

for instance S(1) is the front of chromosome 1/* f=1; */ front 1/* while(max(S)==N+1) for all the chromosomes with S>=f if (cost of chromosome (i) <= (cost of chromosome (j)

or (time of chromosome (i) <= (time of chromosome (j) S(chromosome(i)) f end end f=f+1; end Sort all chromosomes based on S Output: new population

FIGURE 6: PSEUDO-CODE OF FITNESS EVALUATION PART OF PROPOSED ALGORITHMS SELECTION PART



www.elvedit.com 66

IV. DISCUSSION Here, the numerical results of using RB-GA are presented, and also standard GA is used to

proof that the solutions quality obtained using RB-GA is good. As it was explained earlier, instead of having only one solution, there is a set of solution for every case. The proposed algorithm would be able to obtain a PARETO front based on defined selection part which it had already been explained in Figure 6. The solutions exist in PARETO front are the mentioned solution set that should be considered. Every case result presents the non- dominated solutions in PARETO front including transportation cost and product delivery time. For every customer there is a set option of the best delivery route as long as there is a set solution including optimum total logistics cost and optimum total product delivery time. Three problem cases will be examined using GA and RB-GA as shown in following tables:

TABLE 1: PROBLEM CASE 1 WITH THE OBTAINED PARETO SOLUTION

Problem Case # 1: I=2 , J=2 , K=2 , L=20 , Maximum Generation =200

GA RB-GA Cost Time Cost Time 3507 391 1211 224 3545 387 1964 127


Problem Case # 2: I=3 , J=5 , K=7 , L=100 , Maximum Generation =2000

GA RB-GA Cost Time Cost Time

169263 17918 7186 1632 168993 17989 14012 1089


Problem Case # 3: I= 4 , J= 6 , K= 9 , L= 150 , Maximum Generation =4000

GA RB-GA Cost Time Cost Time

266145 25932 31965 3395 264688 26022 35330 3197

- - 35020 3197



The data used in this research was generated by the authors as the problem case. Diverse 3 problem cases have been created by this research to implement the proposed algorithm and compare the obtained solutions. Hardware platforms employed by the researcher were a 2 GHz processor intel core 2 duo with 1GB memory and running windows 7 professinal. It is noted that the scope of this research did not include establishing nessesary conditions to hardware requirements. The mathematical model of fMLN was translated into program written in Matlab version 7, 2009. The problem cases were kept to use for differenet proposed algorithm implementation and the obtained results were displayed in a proper manner in Matlab.

V. CONCLUSION In this paper, the bi-criteria multi source single product fMLN model that considers the

transportation cost and time was formulated. Subsequently, the proposed solution RB-GA which has been explained comprehensively was further developed. To solve multi objectives problems, PARETO solution is needed where a set solution was presented instead of one. The definition of PARETO optimality and non-dominated solution was explained in this paper as well. Additionally, the Pseudo-code of proposed selection part of mentioned algorithm was presented. Lastly, the numerical experiment was explained where it is proven that the obtained solutions using RB-GA are more preferable than the obtained solutions using GA.

The research work presented in this thesis has opened a new line of research in which a number of avenues of future work remain to be investigated. Although there was attempt to solve bi-criteria flexible multistage logistics network problem in this research, however there is still possibility to solve some more variants of fMLN problems by changing the current assumptions of fMLN model and it might be needed to propose some new techniques such as combinatorial algorithms or hybrid algorithms to solve certain specific defined problems. Besides, there is still possibility to propose new methods to decrease the running time of algorithm. Therefore, the following research works could be recommended to enhance the contributions of this area:

1- To solve the fMLN problems when the product price is not fix for every condition of the network. For instance, two types of product price are different when the customer wants to order one by one or wants to order together.

2- To solve multi objectives fMLN problem when there are more than two objectives, such as minimizing the transportation cost, minimizing the inventory cost, minimizing the product delivery time, minimizing the warehousing cost and maximizing the customer satisfaction level simultaneously.

3- To develop the algorithm for constraints handling in a better way for the above mentioned problems



www.elvedit.com 68

References

[1] Alim, F., Ivanov, K. (2005). Heuristic Rules Embedded Genetic Algorithm to Solve In-Core Fuel Management Optimization Problem, GECCO’05, Washington, DC, USA.

[2] Altiparmak, F., Gen, M., Lin, L. & Karaoglan, I. (2007). A steady- state genetic algorithm for multi-product supply chain network design, Computers and Industrial Engineering.

[3] Altiparmak, F., Gen, M., Lin, L. and Paksoy, T. (2006). A genetic algorithm approach for multi-objective optimization of supply chain networks. Computers & Industrial Engineering, 51(1), 197–216.

[4] Branke, J., Deb, K., Miettinen, K., and Roman, S. (2008). Multi objective Optimization: Interactive and Evolutionary Approaches, Springer-Verlag Berlin Heidelberg Printed in Germany

[5] Costa, A., Celano, G., Fichera, S. and Trovato, E. (2010). A new efficient encoding/decoding procedure for the design of a supply chain network with genetic algorithms, Computers & Industrial Engineering 59-986–999.

[6] Craenen, B.G.W., Eiben, A.E. Marchiori, E. (2002). Solving Constraint Satisfaction Problems with Heuristic-based Evolutionary Algorithms - ieeexplore.ieee.org.

[7] Gen, M. (2006). Evolutionary technique for logistics network design: state of the art survey. International intelligence logistics system conference.

[8] Gen, M. and Syarif, A. (2005). Hybrid genetic algorithm for multi-time period production / distribution planning, Computers & Industrial Engineering, 48(4), 799–809.

[9] Gen, M., & Cheng, R. (2000). Genetic algorithms and engineering optimization. New York: John Wiley and Sons.

[10] Gen, M., Altiparamk, F. and Lin, L. (2006). A genetic algorithm for two-stage transportation problem using priority-based encoding, OR Spectrum, 28(3), 337–354.

[11] Gen, M., Cheng, C. and Lin, L. (2008). Network Models and Optimization, Multiobjective Genetic Algorithm Approach, Springer-Verlag London Limited.

[12] Lee, J. E., Rhee, K. G. & Gen, M. (2007). Designing a reverse logistics network by priority based genetic algorithm, Proceeding of International Conference on Intelligent Manufacturing Logistics System, 158–163.

[13] Li, Y. Z., Gen, M. & Ida, K. (1998). Improved genetic algorithm for solving multi-objective solid transportation problem with fuzzy number, Japanese Journal of Fuzzy Theory and Systems, 4(3), 220–229.

[14] Lin, L., Gen, M. and Wang, X. (2007). A Hybrid Genetic Algorithm for Logistics Network Design with Flexible Multistage Model.

[15] Michalewicz, Z., Vignaux, G. A. & Hobbs, M. (1991). A non-standard genetic algorithm for the nonlinear transportation problem, ORSA Journal on Computing, 3(4), 307–316.



[16] Rajabalipour, H., Ishak Desa, M. and Wibowo, A. (2011). A flexible three-level logistic network design considering cost and time criteria with a multi-objective evolutionary algorithm, J Intell Manuf DOI 10.1007/s10845-011-0584-7.

[17] Sheta, A. and Turabieh, H. (2006). A Comparsion between Genetic Algorithms and Sequential Quadratic Programming in Solving Constrained Optimization Problems, AIML Journal, Volume (6).

[18] Syarif, A. & Gen, M. (2003). Double spanning tree-based genetic algorithm for two stage transportation problem, International Journal of Knowledge-Based Intelligent Engineering System, 7(4).

[19] Vignaux, G. A. & Michalewicz, Z. (1991). A genetic algorithm for the linear transportation problem, IEEE Transactions on Systems, Man, and Cybernetics, 21(2), 445–452.

[20] Yaohua, H., Hui, C.W. (2007). Genetic algorithm based on heuristic rules for high-constrained large-size single-stage multi-product scheduling with parallel units.

[21] Yeniay, O. (2005). Penallty function methods for constrained optimization with genetic algorithms , Mathematical and Computational Applications, Vol. 10, No. 1, pp. 45-56.


Impact of Strategic Management Element in Enhancing Firm’s Sustainable Competitive

Advantage. An Empirical Study of Nigeria’s Manufacturing Sector

Authors

Yahaya Sani Business College, Business Administration Department, Sudan University of Science and Technology

[email protected] Khartoum, Sudan

Abdel-Hafiez Ali Hassaballah Business College, Business Administration Department, Sudan University of Science and Technology

[email protected] Khartoum, Sudan

Abstract

The purpose of this study is to investigate the impact of strategy implementation and control as independent variable in enhancing firm’s sustainable competitive advantage through innovation as the dependent variable in the Nigeria’s manufacturing sector. Data were collected through personal questionnaire from166 manufacturing firms in Nigeria who are members of manufacturing association of Nigeria within North West and North central zones with 70% response rate. The results indicate that there is positive and significant relationship between strategic management elements; implementation and control with sustainable competitive advantage; innovation. According to the result manufacturers in Nigeria fully agree that strategy control is essential when a unique strategy has been implemented so as to successfully enhance sustainable competitive advantage. This study adds Knowledge to the theory and practice of sustainable competitive advantage particularly in Nigeria’s manufacturing firms. Its theoretical and empirical significance adds more insight on the previous empirical studies in the field that is to say it gives guidelines to manufacturers in Nigeria on the impact of strategic management approaches on sustainable competitive advantage. For government and firms, the study provides avenue of enhancing sustainable competitive advantage in Nigeria and Africa as a whole since the phenomena is general.

Key Words

Strategy, Competitive Advantage, Sustainable Competitive Advantage and Innovation.

Impact of Strategic Management Element in Enhancing Firm’s Sustainable Competitive Advantage. An Empirical Study of Nigeria’s Manufacturing Sector Yahaya Sani and Abdel-Hafiez Ali Hassaballah


I. INTRODUCTION The concept of free-trade has greatly affected Africa’s economic development in recent past.

According to Tahir,( 2010) Nigeria, which is the most popular nation on the African continent, is highly endowed with lot of human and natural resources , which if adequately harnessed, can turn around not only its economy but the entire economy of Africa. Government regulators are handicapping strategic planning in Nigeria through the globalization policy as too many foreign goods in its market; this has not been possible because Nigeria has allowed itself to be used for all sorts of imported goods from foreign industries and Asian Tigers in the name of globalization. Consequently, this has greatly affected the capacity utilization of various firms of the Nigerian manufacturing sector. Similar view was reorganized by Dembele (1998); Sagagi (2004); Aluko, Akinola and Fataku (2004).

Another serious problem is that if not all greater Nigerian’s income (national income) is from the sale of crude oil and its allied, which is oscillating from one ill to another, this is posing difficulty in applying strategic management principles by manufacturers as manufacturing firms are facing neglect from the regulators. However, it is a thing of concern that even the oil which Nigeria produces, part of it is refined abroad and imported back to the country to meet-up local consumption, because the country’s refineries have over the years been operating below capacity utilization (Daily Trust, 2010).

The situation becomes more aggravated due to Nigerians preference for foreign good (Alukoet al. 2004; Ajayi, 1990). There are few researches on strategic management in emergent markets i.e. developing economy. (Hussam and Hussien (2007), as such Manufacturers in Nigeria do not apply properly strategic management concepts for future development hence this study to turn around the minds of regulators and manufacturers in Nigeria to focus on competitive advantage and push towards sustaining it.

This study contributes to the body of strategic management by empirically investigating the relationship between strategic management elements and sustainable competitive advantage in Nigeria’s manufacturing firms. It also adds to the existing knowledge by applying the resource capability in form of competitive advantage as the mediating variable and environmental factors as moderating variable, all in the name of encouraging the country (Nigeria) to make concerted efforts in the area of trade and manufacturing so as to improve the living standard of its citizens, diversify its economy, enhance revenue generations and survival when the oil reserves are exhausted, and finally to help in achieving its vision 20:2020 of becoming a developed economy.

Modern executive must respond to the challenges posed by the firm’s immediate and remote external environment and often compelled to subordinate the demand of the firm’s internal activities and external environment to the multiple and inconsistent requirements of its stake holders. Thompson et al. (2005) Positioned that generally a company’s strategy should be aimed at providing a product or service that is distinctive from what competitors are offering or developing competitive capabilities that rivals can’t quite match.



www.elvedit.com 72

Manager’s ability to separate powerful strategy from an ordinary or weak one is their ability to forge series of moves, both in the market place and internally, that make the company distinctive as a reason for buyers to prefer its product and or services and produce a sustainable competitive advantage over rivals. Thampson et al. further position that without competitive advantage a company’s risk of being beaten by stronger rivals hence to set strategy that put them apart from rivals in the name of achieving sustainable competitive advantage and performance.

Firms should engage in proper strategic management that enables them to set decisions and actions that result in the formulation, implementation and evaluation of plans to enhance achievement of company’s objectives (Pearce and Robinson, 2003; Snell 2001). Strategic management elements namely: formulation, implementation and control, decide the firm’s resources in form of capabilities as elements of competitive advantage namely: value, rareness, inimitability, cost, differentiation and focus. Since several researches indicate that competitive advantage eroded over time due to environmental factors O’Shannassay (2007) among others. It therefore needs to be sustained to a longer period of time particularly in the Nigeria’s firms.

II. LITERATURE REVIEW Very little researches are available in strategic planning and management in developing

countries (Hussam and Hussein 2007), especially in firm’s Sustainable Competitive Advantage. Hussam and Raef (2010), conclude that majority of researches have shown strong and inconsistent bias toward western context organization, manly in the United States and Western Europe. The developing nation firm’s environment, management characteristics and strategies are different from developed countries, hence the assumption of Hussam and Hussein (2007) that research in strategic management particularly sustainable competitive advantage may be different from that of the developed countries as they do operate in different environment.

A company strategy is viewed as work in progress since most of the times company’s strategy emerges in bits and pieces as a the result of changing circumstances, deliberations on management designs and on gong management actions (Formulation and implementation of strategies) to fine tune their pieces of strategy and to adjust to certain strategy elements in response to unfolding conditions (Thampson et al. 2005). Nonetheless, on occasions, fine -tuning the existing strategy is not enough and major strategy shifts are called for, such as when a strategy is clearly failing, when a market condition or buyers preferences change significantly and new opportunity arises when competitors do something unexpected or when important technological breakthrough occurs.

Despite the criticisms of the Resource Base View on inconsistent findings towards, organizations mainly in U.S.A and Western Europe, according to Hussam and Raef (2010), the study on Strategic planning-firm performance linkage: empirical investigation from an emergent market perspective found that little research is available that investigated the relationship between strategic management and planning and firm performance in other developing and emergent market. Therefore, bringing new data sets from those markets will provide valuable information to answer the question of whether a similar pattern of this relationship prevails



across various contexts. No strategy, no matter how well formulated can achieve longer term success if it is not properly implemented. (Schermerhorne, 2001) organizations must therefore have effective strategic control if they are to successfully implement their strategy. (Dess, and Lumpkin 2003). Formulating the appropriate strategy is not enough, strategic managers also must ensure that new strategies are implemented effectively and efficiently, as organizations recently have been paying more attention to implementation.

According to Snell (2001) clever technique and a good plan do not guarantee success. He further explains that strategy must be supported by decisions regarding the appropriate organizational structure, technology, Human resource, reward system, internal system, organizational culture and leadership style. He also explores that many organizations are extending the more participative strategic management process to implementation. Control from the views of Snell (2001) is characterized as “Steering”. Ordinarily, a good deal of time lapses between the initial implementation of a strategy and achievement of its intended results. During the time, investments are made and numerous projects and actions are undertaken to implement strategy. Snell further expresses at that time, changes are taking place in both the environmental situations and the firm’s internal situation. So, strategic control is necessary to monitor the firm’s activities during the period i.e. the firm’s actions and direction in response to the development and changes.

Porter (1980) stands that firms must keep on innovation as its revenue stream is consistently espoused to new competitors, substitute product and so forth. Furthermore, Hoffmann (2000) positions that; the fundamental basis of long-run success of a firm is the achievement and maintaining of a sustainable competitive advantage. According to Jeroen, Spender and Aard (2010) firms are not passive, a competitive advantage can be sustained only at the dynamic level through advantageous dynamic capability or organizational learning enabling the firm to adopt faster than its competitor.

The study further stands that sustainable competitive advantage directs management attention on the dynamics that support i.e., emphasizing the sustainability, looking for practical ways of beating the market natural timing, quickening innovation or slow imitation. Jeroen et al. (2010) further conclude that in a dynamic environment firms cannot derive sustainable competitive advantage from a static set of resources.

III. FRAMEWORK Based on the literature review, the integrative framework of this study is on resource based

view to determine impact of strategy implementation and control in attaining sustainable competitive advantage through innovation. The study examines the impact of strategic management element as independent variable consisting of two constructs (implementation and control) on sustainable competitive advantage with one construct innovation as the dependent variable (See Figure 1).



www.elvedit.com 74

Strategic Management Elements Sustainable Competitive Advantage

FIGURE 1: CONCEPTUAL FRAMEWORK SOURCE: CREATED BY THE STUDY

IV. HYPOTHESIS In this study two main hypotheses were developed to test the relationship between strategic

management elements and sustainable competitive advantage i.e. the relationship between implementation and innovation and the relationship between control and innovation.

According to According to Pearce and Robinson (2007) strategic formulation guides executive in defining the business their firm is in, the ends it seeks, and the means it will use to accomplish those ends. They further conceptualized that a firm’s competitive plan of action; strategy formulation combine a future orientation perspective with concern for the firm’s internal and external environment (Value). Pearce and Robinson (2007) position that whether a firm is developing a new business or reformulating direction for an ongoing business, it must determine the basic goals and philosophy that will shape its structural posture, that set a firm apart from other firms of its type and identify the scope of its operation in product and market terms, in essence sustainable competitive advantage.

Strategic management consists of the analyses, decisions and actions an organization takes in order to create and sustain competitive advantage. (Dess and Lumpkin 2003) The definition entails three ongoing processes: Analysis, Decision and Action, and according to Pearce and Robinson (2003) Formulation, Implementation and Control. Burden and Proctor (2000) found out that meeting customers’ needs on time, every time is a significant route to achieving sustainable competitive advantage and innovation is a tool that organization should be using to succeed.

Amit and Schoemaker (1993) in their study found that management skill of Formulation, implementation and control of strategy in itself is a source of sustainable competitive advantage. Based on the discussion above the following hypothesis is generated:

H1 There is positive relationship between implementation and innovation.

Dess and Lumpkin (2003); Pearce and Robinson (2003) state that strategic management explain to managers why some firms outperform others and how firm can obtain advantage sustainable over a lengthy period of time. This also means focusing on two fundamental questions: how should we compete for advantage, and how should we make advantage sustainable? Strategic management is an ongoing process that evaluate and control the business and the industries in which the company involves, assesses its competitor and set goals and strategies to meet all existing and potential competitors, and then reassesses each strategy annually or quarterly (regularly) to determine how it has been implemented and whether it has

. Implementation . Control

Innovation



succeeded or need replacement by a new strategy to meet changed circumstances.

Alderson, (1965) was one of the first to recognize that firms should strive for unique characteristics in order to distinguish themselves from competitors in the eyes of costumers. Hall (1980) and Henderson (1983) solidified the need for firms to possess unique advantages in relation to competitors if they are to survive. This arguments form the basis for achieving sustainable competitive advantage.

Porter (1980) stands that firms must keep on innovation as its revenue stream is consistently espoused to new competitors, substitute product and so forth. Furthermore, Hoffmann (2000) confirms that the fundamental basis of long-run success of a firm is the achievement and maintaining of a sustainable competitive advantage. The study concludes that knowing that given the intense nature of competitor today , firms must be more innovative and environment conscious in their strategic planning than just lowering price.

Fahy (2000) in the study capitalized that sustainability does not refer to a particular period of calendar time, nor does it imply that advantage persists indefinitely. It erodes overtime hence strategy control. Barney et al. (2001) and Helfat et al. (2007) conclude that to develop competitive advantage into sustainable competitive advantage is harder to do, because the firm must possess value creating things or capabilities that cannot be made redundant by dynamism in the environment. Based on the discussion above the following hypothesis is generated:

H2 There is positive relationship between strategy control and innovation.

V. RESEARCH DESIGN The study considers a survey method being a popular and common strategy in business

research, because it allows for the collection of large amount of data from sizable population in a highly economical way. Therefore, this research considers questionnaire tool while date collections. Sunders et al. (2007) are of the opinion that research project for academic courses are time constrained. Therefore, in this study due to time management a cross-sectional strategy is employed - a study in which a group of individuals are composed into one large sample and studied only at a single point of time.

This research decides a cross-sectional approach as the dominant method in marketing research and the questionnaire survey approach sounds most appropriate means to collect data from the manufacturing firms in Nigeria.

VI. DATA ANALYSIS A total of 166 questionnaires were distributed to respondents through personal administered

questionnaire. While total of 116 questionnaires were collected. The overall response rate was 70%, to ensure the goodness of measurement exploratory factor analysis (principal component analysis) was conducted on strategic management elements and innovation. In addition, Reliability test (Cronbach’s alpha) was done to measure the internal consistency of the items



www.elvedit.com 76

used on the questionnaire. These two methods were very important to assess the goodness of the measures (Sekaran 2003). Correlation test was conducted to measure the relationship among the variable, and regression analysis was also conducted in the study in order to test the relevance of the hypotheses.

A. Factor and Reliability Analysis on Strategic Management Elements

The original questionnaire had three dimension, eight items for Formulation, eight for Implementation and seven measuring Control. The result of factor analysis indicates that two factor i.e. Implementation/ formulation and Control as suitable. The two dimensions is not a set back as formulation is merged with implementation. Barney (1991) affirms that firm is said to have competitive advantage when it is implementing a value creating strategy and not currently being implemented by present or possible future company. So, Implementation/formulation and Control stand.

Based on factor analysis result, the measurement of (KMO) was .71 and the Bartlett test of sphericity was significant, both indicating there is sufficient inter correlation among the factors. The result was achieved after deleting several items, for insufficient correlation i.e. was found to have communalities less than 0.50. The table 1 shows the result of analysis. The two loading factors ranging from .72 to .60, factor one has three questions and also three questions from factor two. The 2 factors cumulatively captured about 58% of the total variance in the data. All items have factor loading above 0.50 with Eigen value of 4.56 and 2.26.

The factors are subject to varimax rotation and the name of the factors was renamed factor 1 implementation/formulation and factor 2 control. The reliability value (Cronbach’s alpha) for Implementation was (.71) and Control (.71). All assumptions were satisfactory fulfilled. All the remaining items had more than recommended value of at least 0.50 in MSA with KMO value of .71 (above the recommended minimum level of (.60) and Bartlett’s test of sphericity is significant (p<.01) and Eigenvalue above 1. This shows that according to manufactures in Nigeria strategic management elements are two recognizing the implementation and formulation as one.

TABLE 1: FACTOR AND RELIABILITY ANALYSIS ON STRATEGIC MANAGEMENT ELEMENTS

Variable and Question Items Factor loading F1 F2

Implementation / Formulation My company employ’s good information system .72 .026 My firm observes its budget head to head .62 .158 My firm adopts sound performance indicator Control .72 .022 May firm takes cognizance of changing market test .272 .60 May company adopts good leadership style .084 .70 My company has good organizational structure .022 .61 % of Variance explain 28.43 30.00 Eigenvalue 4.56 2.26 Reliability .71 .71

Variables loaded significantly on factor with Coefficient of at least 0.5, * Items deleted due to high cross loading



B. Factor and Reliability Analysis on Sustainable Competitive Advantage Initially the questionnaire had three dimension i.e. Innovation, training and development, and

Reward. Innovation sounds suitable for the study, factor analysis was conducted on innovation only and the result from table 1 shows innovation with 4 items as one factor loading, with (MSN) value above 0.50, KMO was (.71) and Bartlett sphericity test was significant. The one factor cumulatively captured 62% variance of the date and Eigenvalue 2.50. The original name of the factor remains. The reliability (Cronbach’s alpha) was (.80). All assumptions were satisfactory fulfilled. All items had more than recommended value of at least 0.50 in MSA with KMO value of .71 (above the recommended minimum level of 0.60), and Bartlett’s test of sphericity is significant (p<.01 Eigenvalue above 1. This indicates that manufacturers in Nigeria recognize innovation as a means of enhancing firm’s sustainable competitive advantage.

TABLE 2: FACTOR AND RELIABILITY ANALYSIS ON SUSTAINABLE COMPETITIVE ADVANTAGE Variable and Question Items Factor Loading F1 Innovation My firm engages in resources preemption .93 My firm benefits from technological Leadership .92 My firm innovates first then others follow in the industry .72 My firm enjoys patent right .51 % of Variance explain 62% Eigenvalue 2.50 Reliability .80

Variables loaded significantly on factor with Coefficient of at least 0.5, * Items deleted due to high cross loading

TABLE 3: INTER CORRELATION OF VARIABLES Variable Mean S/Deviation Implementation Control Innovation Implementation 4.02 .519 1.00 .381*** .304*** Control 4.31 .492 .381*** 1.00 .390*** Innovation 3.78 .827 .304*** .390*** 1.00

***p<0.01 and **p<0.05

Table 3 above shows the result of the inter correlations among the variables. The tables indicated that the mean value for both variables is above average, indicating that they are positively and sufficiently correlated with each other. The table shows that implementation positively and sufficiently correlates with control (r=.381***p-value<0.01) and correlates also with innovation (r=.304**p-value<0.05). The test also indicates that control positively and significantly correlates with innovation (r=.390***p-value<0.01). Therefore, both independents and the dependents variable of this study are sufficiently correlated.

VII. RESULTS AND DISCUSSION The table 2 above shows the result from hierarchical regression between Strategic management

elements (Implementation and Control) and Sustainable competitive advantage (innovation). The strategic management elements variable cumulatively contributed 62% of the variance in



www.elvedit.com 78

innovation. The table further shows positive and significant relationship between the variables, hence H1 (Implementation and Innovation), H2 (Control and Innovation) were accepted.

TABLE 4: MULTIPLE REGRESSIONS: IMPACT OF STRATEGIC MANAGEMENT ELEMENTS ON SUSTAINABLE COMPETITIVE ADVANTAGE (INNOVATION) (BETA COEFFICIENT)

Variables Innovation Implementation Control

.162* .309**

R2 Adjusted R2 ∆ R² F change

.160

.145

.145 10.758**

**p<0.01 *p<0.50

Therefore, the regression coefficient from the table above indicates that the independent variable control is the most important in explaining the variance in Innovation with (β=.31) then Implementation with (β=.16) and the F-change is significant.

The finding from this study explains that there is positive and significant relationship between strategic management element (implementation and control) with sustainable competitive advantage (innovation). The result indicates positive and significant relationship between implementation and innovation. The findings from past studies equally indicate similar relationship (e.g. Pearce and Robinson 2007; Barney (1991); Newbert (2008); Mabey et al. (1998); Lahteenmaki, et al (1998); Aniloui (1999) state that whether a firm is developing a new business or reformulating direction for an ongoing business, it must determine and implement the basic goals and philosophy that will be its structural posture, that equally set it apart from other firms and identified the scope of its operation.

In essence (innovation) sustainable competitive advantage. Likewise, previous studies from Fahy (2000); Burden and Proctor (2000); Gupta and McDaniel (2002) Aniloui (1999) are of the opinion that, there is some agreement at least on one point i.e. a link between firm’s strategy and utilization of human resources, indeed people management can be a key sources of sustainable competitive advantage. This positive relationship indicates the commitments of manufactures in Nigeria towards innovation through implementing strategies that enhance sustainable competitive advantage.

According to the findings, significant relationship between control and sustainable competitive advantage (innovation) was material. Past studies acknowledge similar relationship (e. g Dess and Lumpkin 2003; Peace and Robinson 2003; Amit and Schoemaker 1993) all agree that strategic management consists of implementation and control of strategies that an organization undertakes in order to create sustainable competitive advantage. Findings from past researches from Burden and Proctor (2000); Gupta and McDaniel (2002) equally support positive relationship between control and sustainable competitive advantage, in the sense that, meeting customer need on time and every time is a significant tool or route to achieving sustainable competitive advantage, through training and knowledge management (innovation). Furthermore,



Porter (1980); Hoffmann (2000); equally posit that knowing the intense nature of competition today, firms must be innovative in their strategic planning constantly as their revenue stream is exposed to new competitors. Previous researches from Fahy (2000); Barney (1991, 1994) and Aniloui (1999) all display positive relation between the variables, in essence implementation and control. This positive and significant relationship shows that manufactures in Nigeria fully agree that strategy control is essential when a unique strategy has been implemented so as to successfully enhance sustainable competitive advantage.

VIII. THEORETICAL AND MANAGERIAL IMPLICATIONS The theoretical implication isolated in this study reveals that strategic management elements

(Implementation and Control) are very important sources of resource creation and generation of firm’s capabilities in form value, rareness and inimitability (Competitive advantage). This result is consistent with Resource Base View which states that a firm is said have competitive advantage when it is implementing a value creating strategy not implemented simultaneously by any current or potential player. This study of manufacturing firms in Nigeria contributes and supports the theory and various studies carried out by several scholars in the area of resource base view, as well its contributions to the Nigeria’s firms and theory as well as to emerging economies.

The findings also provide evidence of relationship between elements of competitive advantage (Value, Rareness and Inimitability) and sustainable competitive advantage (Innovation). This was found in Barney (1991); Newbert (2008) among others who are advocates of resource base view, that firm must identify and implement resource- based strategy (Value, Rareness and Inimitability) to sustain competitive advantage in producing product with more benefits in form of unique features (innovation). This study also supports the view of sustainable competitive advantages as its finding contributes to the area of sustainable competitive advantage

Furthermore, the study provides interesting insight for understanding the direct link between strategic management elements (implementation and Control) with sustainable competitive advantage (Innovation). This positive relationship encourages the view of Peace and Robinson (2007) that firms must determine and implement basic goals and philosophy that will be its structural posture at the same time set it apart from other firms, in essence (innovation) sustainable competitive advantage. The result indicates that implementing and control of value creating and inimitable strategy means innovation which enhances sustainable competitive advantage. The study contributes in the literature of resource base view and sustainable competitive advantage particularly in the emerging economies Nigeria inclusive.

The result from the findings of this study should awaken managers to the fact that adequate utilization of strategic management elements (Formulation, Implementation and Control) is essential in resources creation, which results in competitive advantage and when managed properly sustains the advantage. Manufacturers in Nigeria should equally put adequate concern on strategic formulation as they do to Implementation and control. Manufacturers in Nigeria



www.elvedit.com 80

according to the findings of this study conceded and view strategy formulation as part of strategy implementation which shows discrepancy with previous studies.

IX. LIMITATIONS AND FUTURE RESEARCH There are a number of issues that could be addressed in future researches, aiming at

developing a kind of comprehensive understanding of the impact of strategic management elements in enhancing sustainable competitive advantage in Nigeria’s manufacturing firms. This stands a clear limitation for the study due to its inability to include service industry too. Hence there should be future researches to consider other areas say service firms among others with similar framework.

The study is supposed to cover the six geo-political zones in Nigeria, but due to the time framework and financial implications, the study sampled two important zones i.e. North West and North Central. Therefore, a similar study to be conducted in other zones particularly south west where there are concentrations of company equally. Further studies can use the strategic orientation (cost, focus and differentiation) using the same framework as the mediator between independent variable and the dependent variable to compare with the findings from this study. Similarly there should be future study to include all the elements of completive advantage and compare the findings.

X. CONCLUSION The main aim of this study is to examine the impact of strategic management elements

(implementation and control) in enhancing sustainable competitive advantage (Innovation) in the Nigeria’s manufacturing firms. The research findings affirmed that resources based view theory has impact on firm’s competitive advantage as well as sustaining the advantage. This has been seen from various past researches, and the findings from this study also supportes the theory. Furthermore, the study on enhancing sustainable competitive advantage in the Nigeria’s manufacturing firms found that proper strategy implementation and control has impact on firm’s competitive advantage as well as sustaining the advantage further. This has been seen from various past researches and the findings from this study also support the theory.

The study encourages managers and government in the utilization of the concept of strategic management in the name of sustaining competitive advantage. It also solicits future studies to be conduct to see the impact of strategy formulation as it merged with strategy implementation as well as similar study be conducted in other sectors of the economy with the same framework for instance service sector among others. The findings provide empirical support for the theoretical framework, demonstrating the fact that the study has sufficiently addressed the research hypotheses. The study also highlights the implication, limitations and suggestions for future research.

REFERENCES [1] Ajayi, I. (1990). Nigeria: The Trapped Economy. Ibadan- Nigeria: Heinemann Educational Books Ltd.



[2] Alderson, W. (1965), Dynamic Marketing Behavior: A Functionalist Theory of Marketing. Homewood, IL: Richard D. Irwin, Inc.

[3] Aluko, M.A.O., Akinola, G.O. and Fatakun, S. (2004). “Globalization and the Manufacturing Sector: a

Study of Some Selected Textiles Firms in Nigeria”. Journal of social science, Vol. 9 No. 2, pp 122-127. [4] Amint, R. and Schoemaker, P. (1993), “Strategic assets and organizational rent”, Strategic

Management Journal, 14:33-36 [5] Analoui, F. (1999), Effective Human resources development: a challenge for developing Countries

Aldershot: Ashgate. [6] Barney, J. B. (1991a), “Firm resources and sustained competitive advantage”, Journal of Management,

17: 99-120. [7] Barney, J. B. (1994). Bringing managers back in: A resource-based analysis of the role of Managers in

creating and sustaining competitive advantaged for firms in Does Management matters? On Competency and competitive advantage. The 1994 Crawford lectures: 1-36. Lund University, Institute of Economic Research.

[8] Barney, J. B.( 2001), “Is the resource-based view a useful perspective for strategic management

research? Yes. Academy of management Review, 26: 41-56 [9] Burden, R. and Practor, T. (2000), “Creating Sustainable competitive advantage trough training”, Team

performance management, vol.6, No. 5/6, pp.90-97 [10]Daily Trust (2010). Editorial. Vol. 18, No. 94, April 24, p 1. [11] Dess. G. G. and Lumpkin, G. T. (2003), Strategic Management. McGraw – Hill Irwin New York. [12]Fahy, J. (2000), “The resource-based view of the firm: some stumbling blocks on the road to

Understanding sustainable competitive advantage”, Journal of European industrial training, pp.94-104 [13] Gupta, A, and McDaniel, J, (2002), “Creating competitive advantage by effectively managing

Knowledge: A framework for knowledge management” Journal of knowledge Management practice, [14] Hall, W. K. (1980), "Survival Strategies in a Hostile Environment." Harvard Business Review 58

(September-October): 75-85. [15] Henderson, B. (1983), "The Anatomy of Competition." Journal of Marketing 47 (spring): 7-11. [16] Hussam, A.A. and Raef, T. H. (2010), “Strategic planning-firm performance linkage: empirical

investigation from an emergent market perspective”, Advance in competitiveness Research Annual, 2007.

[17] Helfat, C.E., Finkelstein, S., Mitchell, W., Peteraf, M.A., Singh, H., Teece, D. J., and Winter, S. G.

(2007), Dynamic capabilities: Understanding strategic change in organizations. New York: John Wiley. [18] Hoffmann, P. (2000) “An examination of the “Sustainable competitive Advantage Concept: Past,

present and future” .http/[email protected] [19] Jeroen, K., Spender, J. C., and Aard, G. (2010), “The Resource-Based View: A Review of Its Critiques”

Journal of Management 36:349.



www.elvedit.com 82

[20] Lahteenmaki, S., Storey, J, and Vanhala, S. (1998). “Human resources management and company

performance: The use of measurement and influence of economic cycles” Human resources management journal, vol. 8, No. 2 pp.51-65

[21] Mabey, C., Salman, G. and Storey, J. (1998), Strategic human resource management, London: SAGE

Publications [22] Newbert, S.L. (2008) “Value, rareness, competitive advantage, and performance: A conceptual-level

empirical investigation of the resource-based view of the firm” Strategic management Journal, 29(7),745-768.

[23] O’Shannassay, T (2008) “Sustainable competitive advantage or temporary competitive advantage:

Improving understanding of an important strategy construct”, Journal of strategy and management vol.1 No.2. Pp.168-180.

[24] Pearce, J. A., and Robinson, R. B. (2003), Strategic Management: Formulation, Implementation, and

Control. McGraw – Hill Irwin. New York. [25] Pearce, J. A., and Robinson, R. B. (2007), Strategic Management: Formulation, Implementation, and

Control. McGraw – Hill Irwin. New York. [26] Sagagi, M.S. (2004). Globalization, International Competition and Corporate Strategy: An Analysis of

Textile Industry in Northern Nigeria. A PhD. Thesis submitted to Usman Danfodio University, Sokoto-Nigeria (Unpublished).

[27] Sekaran, U. (2000). Research in methods for business: a skill-building approach. John Wiley & Sons

[28] Schemerhorne, J. F. (2001), Management. 6th Edition. John Wiley & Sons. New York. [29] Snell, B. (2002), Management. 5th Edition. McGraw – Hill International edition. [30] Tahir, K.B. (2010). Globalization and Nigeria’s economic development: Rethinking Nigeria’s trade and

industrial policy. ABGA International conference Malaysia. [31] Thompson, AA., Strickland, AJ., and Gomble, JE. (2005). Crafting and executing strategy: The for

competitive advantage. 14th edition. Mc Grew – Hill international edition New York.


A Co-modal Transport Information System in

a Distributed Environment

Authors

Zhanjun Wang LAGIS, Ecole Centrale de Lille

[email protected] Villeneuve d’Ascq, 59651, France

Khaled Mesghouni LAGIS, Ecole Centrale de Lille


Slim Hammadi LAGIS, Ecole Centrale de Lille


Abstract

This paper is aimed at presenting a transport information system that is dedicated to the co-modal transportation services. The problem is formulated with a three layers model and this work concentrates on the second layer — the assignment of the vehicles on each section of the itineraries. In terms of cost, travel time and other criteria, the optimization for choosing the best route for each request is implemented with Evolutionary Algorithms (EAs) and local search algorithm for the allocation of limited transportation resource. A special encoding method is developed to adapt the concerned problem and the operators for EAs are also detailed. With the aggregation approach, the fitness function is defined for EAs. According to the size of requests and the characteristics of the problem, an appropriate algorithm will be selected. With respect to users’ preferences and availability of vehicles, the simulation is provided in this contribution to illustrate the proposed method.

Key Words

Assignment, co-modal transport, distributed network, EAs, optimization.



www.elvedit.com 84

I. INTRODUCTION For the economical and environmental reasons, some new modes of transportation emerge and

get more and more popular in recent years, like the carpooling (e.g., http://www.covoiturage.fr) and free-use car (e.g., AutoLib in Paris). In 2006, the European commission introduced a new concept: co-modality [1] that refers to a use of different modes on their own and in combination in seeking for the aim of an optimal and sustainable utilization of resources. With this notion, it means finding an optimum searching in the relevant domain of the various transports (including individual and public transport) and of their combination, in a way where travel cost, traveling time, distance, environment impact, comfort conditions, quality of service etc. are taken into account. In a co-modal transport information system, travel information about various transportation, like the best routing and schedule in the above optimum way is provided to the users. Its decentralized features of its data characterize a distributed information system. Generally, only the subsystems take full control of their own data. When necessary, the detailed information about one route such as the schedule and the availability of the cars will be demanded through networks, and meanwhile, the subsystems offer their local information. One entire journey may involve several transport operators, then different subsystems. The main system is in charge of finding out the global information throughout the subsystems that process the local information. The distributed information system avoids the maintenance and updating of the data from each information provider, and for the privacies reasons for the companies, it’s more acceptable. Essentially, the departure spot, the destination spot and the related time are indispensable to launch an itinerary query. For advanced options, constrains like the preference of transportation tools and the flexible time window allow the system to return more suitable itineraries. The agent computing paradigm is rapidly emerging as one of the powerful technologies for the development of large-scale distributed systems to deal with the uncertainty in a dynamic environment due to its autonomous, reactive and proactive nature [2] [3]. In this dynamic problem, the subsystems are independent from each other, and the member routes will interact together to formulate a complete itinerary, under the communication protocol. The goal of the paper is to describe a new approach to resolve the time-dependent co-modal problem in the distributed transportation networks. Along with the mathematical notions, some scenarios of simulation are also presented to verify its efficiency. The reminder of this paper is structured as follows: Section II contains research overview, where the state of art in the multi-modal transport, distributed environment transport optimization is emphasized. In Section III, the advantages and the main features are given. In section IV, with the formulation of the problem in a mathematical way, the solution to the problem and the relevant model are described. Section V presents the configurations of performed experiences and the obtained simulation results. In the last Section, conclusions and future works will be presented.

II. RESEARCH OVERVIEW Based on the graph theory, Dijkstra’s algorithm for the shortest route is the most used method,

and lots of variant approaches [4] were developed for different problems and better computation performance. In [5], the research focused on the finding shortest path in urban multimodal transportations networks and minimizing the cost, time and other discommodity related to the paths. A utility function using the aggregation method considers the arcs cost and time weight in

A Co-modal Transport Information System in a Distributed Environment Zhanjun WangANG, Khaled Mesghouni and Slim Hammadi


the meantime the preference of the users associated with all the possible transportation modalities. Specifically, a purely theoretical proposal of possible values for the weighting coefficients is presented, especially the ones about the time and cost. Parameters should be well identified in order to precisely present how these factors affect users’ choices. In [6], Wang and Kampke introduced an algorithm for computing the shortest route in a distributed system during polynomial time. In a distributed system, there are several autonomous subsystems and a central computing center. The subsystems maintain and take full control of their own databases and provide the central computing server with the intersection information for transfers among themselves. In [7], an approach about identifying the fastest itinerary in a time-dependent distributed environment is presented. All the transportation vehicles are well scheduled in the subsystems. The central computer uses the transfer information and the incomplete local information provided by the independent subsystems to search the fastest itinerary from one spot to another across the subsystems. In [8], the research focused on the dynamic carpooling problem in a distributed transportation network.

Despite the great efforts made in this field, the complexity of the time-dependent co-modal transport problems has not been completely addressed. Our work mainly focus on the finding of the appropriate route for the users in condition that the resource like the availability of carpooling, the number of free-use cars are limited and that the cost and time constrains are imposed by the users. The genetic algorithm and the local search algorithm will switch automatically for a better performance. A completed itinerary can be composed of several routes, and there must be transfers between different transport operators. After the possible routes are identified, a process to form an entire itinerary from the departure spot to the arrival spot is executed. A protocol for the negotiation between the agents is also necessary to establish.

III. SYSTEM AIMS AND METHODOLOGY The goal of the system is to establish a co-modal transportation information system with which

the demanded itineraries will be optimized with respect to the criterions and the preferences of the demanders and will be returned within a reasonable limit of time in case of immense quantity of quests. For the optimization process with EAs, a special encoding method is adopted. The optimization that is also a multi-criterion problem will be executed with the aggregation method for the fitness assignment function.

A. The Co-modal Transportation Network

In a multimodal transportation, an origin-destination path is composed of several sections (or routes) and each of them is ensured by one modality (car, train etc.), instead, in a co-modal transportation system, a section in one origin-destination path may be served by more than one modality where both public and private ones are considered. And these modalities compete with each other. After having received the requests from users, the system will find the most suitable modality from each section and then formulate a full path.



www.elvedit.com 86

Let 퐺 = (푉,퐸,푀) denote a directed transportation network, where 푉 = (푣 , … , 푣 ) is the set of nodes representing the relevant spots, 푀 = (푚 , … ,푚 ) is the set of transportation modalities (e.g., train, subway, car pooling and free-use car) [9] and 퐸 = (푒 , … , 푒 ) is the set of the directed arc (also called segment). A segment 푒 ∈ 퐸 connecting two nodes 푣 and 푣 can be determined using 푣 , 푣 mj where 푚 ∈ 푀 represents the relevant transportation modality.

Definition (Multimodal path [9]): In a given multimodal graph G = (V, E, M), a multimodal path (v , v ) is a sequence of edges between a pair of nodes v and v , the segments between these two nodes are represented by ((푣 , 푣 ) , … , (푣 , 푣 ) ), where ∀푖, 푗 ∈ {1, … ,푘}, 푣 ,푣 ∈ 푉, (푣 ,푣 ) ∈퐸,푚 ∈ 푀, and 푖 ≠ 푗 ⇔ 푣 ≠ 푣 .

Definition (Co-modal path [10][11]): In a given directed graph G = (V, E, M), a co-modal path (v , v ) is a sequence of edges between a pair of nodes v and v , the segments between these two nodes are represented by {( v , v

,…,)} , where ( v , v

,…,) means that the segment

(v → v ) can be ensured by any one mode in {m , … , m } . In this paper, the following transportation modalities will be considered in the co-modal transportation network:

• Public modality (e.g., bus, metro, train): for the public transportation modalities, the departure and arrival time are usually predefined and the places available are considered as being enough. In the system, we take these characteristics into account.

• Carpooling: the departure and arrival time are usually predefined and may be changed anytime before departure, the most important factor is the available places; another characteristic is that one traveler can only take a partial itinerary.

• Free-use car: this modality is specially described by the number of available vehicles at the departure spot, besides, the cost of the travel composes of the service subscription fee and the use fee.

B. The Three Layers Model

In this part, a three layers model of the proposed solution for the co-modal transportation problem will be presented. From the users’ itineraries requests, the global optimal itineraries will be obtained [6] [7]. Each itinerary composes of one or more segments that are not necessarily ensured by the same transport operator and the transportation tools. The itineraries may also have segments in common, so the travelers share the segment of the route. In particular, the resource (e.g., carpooling) is limited. To get better allocation of the available resource and to well arrange the travelers’ journeys according to their preferences, a three layers model is implemented. Figure 1 is the proposed three layers model.

The first layer is served as the interface and the identification of global itineraries. The identified itineraries will be decomposed according to the areas, the transport operators and the service modalities. Thus, the decomposed segments and the demanded itineraries will be sent to the second layer and the third layer. In the second layer, the assignment for each traveller will be accomplished in the limit of available source of transportation service and the travelers’ preference. In the second layer, the modality will be chosen for each traveller in each segment. At



last, the assigned segments and the demanded itineraries permit la coalition among the routes to formulate the entire journey. The third layer is the last step to form a whole itinerary from the allocation agents gotten in the second layer. This step will be on the basis of some interaction protocol. The process of getting the segments allied is named in this paper combination of the segments. By the way, each layer contains a process of optimization.

Figure 2 is the entire activity diagram of this information system. The last part that is marked grey is to be worked on in the future. We will first study on the identification of the entire itineraries and the sections along with the possible service that ensures the transportation. The coalition (combination) of the identified sections will be a part of future work.

IV. PROPOSED APPROACH In the dedicated transport information system, the work consists of three steps. First step, it’s

to find the complete itinerary in the distributed environment to the requests. Second step, the itinerary found in the last step is probably composed of several segments of route, and services are provided by different operators, at the same time some of them have alternatives, for example, train, carpooling and free-use car are available to travel from one place to another during the entire journey. In this step, the system will find the optimized distribution transport tools for each traveller in condition that the resource like free-use cars is limited. The third step is to choose and to combine the optimized segments into a complete itinerary with the users constraints and preferences being taken into account.

The topic of this paper is to treat the problem of the second layer of the three layers model. After having gotten the global itineraries in the preceding step, to allocate the available vehicles with respect to the users’ preferences in each segment in a co-modal transportation network is the next step to carry out.

FIGURE I: THE STRUCTURE OF THE THREE LAYERS MODEL

FIGURE II: THE ENTIRE FLOW CHART OF THE INFORMATION SYSTEM



www.elvedit.com 88

In the first step to find the shortest path in a distributed environment, we use the following approach. The use of transfer graph [9] and the approach proposed by [6] allows a rapid computation of a shortest path in a distributed system. Here, we will present the major steps rather than the details of the algorithms. Each operator of transportation is treated as an individual system and the graph that represents its network is individual. There is no central computing server to store the data and the problem becomes to find out the shortest path in the inter-class route graph. The Dijkstra’s is always the core algorithm for the “shortest path”.

A. Mathematical Formulation of the Problem

The problem is defined in the mathematical way as follows:

• At a time 푡 , there are 푁 quasi-simultaneous requests for itineraries during a little interval of time 훥푡. Use 푅(푡) to indicate the set of requests.

• In the distributed co-modal transportation network, 퐼(푡) will be returned as the global optimal itineraries for the requests in 푅(푡).

• ∀푖 ∈ {1,2, … ,푁}, there are 퐼(푡) ∈ 퐼(푡), 퐼(푡) = {푆(푡) , … , 푆(푡) } where 푆(푡) , … , 푆(푡) are segments of 퐼(푡) .

• At the same time, there are segments shared by several itineraries. Let us take 푆(푡) as an example to study. Suppose that the segment 푆(푡) is shared by 퐾 paths in the time window [푡 , 푡 ] noted as 푇.

• To assure the segment 푆(푡) , in the corresponding time window 푆(푡) , there are 푄 places available for car-pooling 푀 , 푄 available free-use cars 푀 , and public transport 푀 that is considered with unlimited places for travelers.

B. The Assignment of the Vehicles

An efficient representation method is employed for the suitable adaptation to the problem. For each segment of the itinerary, we use the following method to indicate the transport modality that the traveler will engage.

We use the matrix for the assignment. In the following, we will explain how this works. For each element of the matrix, it represents the assignment of a user to one transportation tool or one vehicle. Each specific segment is represented by one matrix. The rows of the matrix correspond to the requests demanders whose itineraries may contain this route and the columns are related to the vehicles that ensure the service on this segment. With 푆(푡) referring to the route, 푃 referring to one traveller on route 푆(푡) and 푉referring to one vehicle that ensures service on the segment 푆(푡) , 퐶퐻 denotes to the name of this assignment matrix, 퐶 is the row index and 퐶 is the column index, the element of the matrix is the assignment of the traveller in the following way:

퐶퐻 퐶 ,퐶 (푆(푡) ) =1,if푃isassignedto푉∗, if푃canbeassignedto푉0,if푃can'tbeassignedto푉

(1)



Then, with this method of representation, we will get a matrix of assignment where the columns are the vehicles and the rows are the travelers. As a result, each element of the matrix indicates the assignment relation between the traveler and the vehicle. There are in total three different cases: firstly, the element “1” means that the traveler 푃 will be assigned to the vehicle 푉; secondly, the element “∗” in the matrix means that the traveler 푃 may take the vehicle 푉 as one of his options; at last, the element “0” implies the impossibility for the traveler 푃 to take the vehicle 푉.

So, the size of the matrix, which means the number of the potential travelers on this route as the rows, the number of the vehicles ensuring the service in the same time window as the columns, depends on the requests received by the transport information system simultaneously. It is impossible that a traveller takes more than one transportation tool on the same route without transfer, so he can be assigned to only one vehicle. In the case that some modalities are excluded in the travellers’ preference, the assignment to the relevant vehicles will not occur.

C. Evolutionary Approach

The problem can be treated as an allocation problem in which the limited transportation resource is allocated to the travelers. Some criteria like time and cost will be followed to get the allocation optimized and the travelers’ preferences will be respected. We propose an evolutionary algorithm to accomplish this optimization process. This assignment problem is a combinatory multi-objective problem. The meta-heuristics are often used to solve this type of multi-objective combinatorial optimization problem [12][13]. For this problem, we choose an evolutionary approach.

In the first step of an evolutionary algorithm, it is necessary to establish an encoding method with which the chromosomes are formulated. An efficient encoding method is adopted for the suitable adaptation to the problem. With this encoding, the chromosome is in the form of matrix. To obtain the optimal assignment matrix for the route in terms of travel cost, travel time, travelers’ preference etc., an evolutionary approach is implemented. According to its characteristics, the problem transforms into a combinatory optimization problem. As the sections of route are different from each other, we choose some indicators to describe the sections. In other words, the indicators are called attributes of the route that allow us to execute the optimization process. Traveling with a certain transportation tool in one route, the price paid is the travel cost and the time spent is the travel time. There are also others attributes like total gas emission, comfort conditions etc. To take several criteria into account for the optimization makes the problem become a multi-criterion and multi-objective problem.

Let us consider a multi-objective optimization problem with 푛 objectives:

Maximize푓 (풙),푓 (풙) … ,푓 (풙)

where 푓 (∙),푓 (∙) … ,푓 (∙) are 푛 objectives to be minimized.

When the following inequalities hold between two solutions 풙 and 풚, the solution y is said to dominate the solution 풙:



www.elvedit.com 90

∀푖: 푓 (풙) ≤ 푓 (풚)and∃푗: 푓 (풙) < 푓 (풚)

If a solution is not dominated by any other solutions of the multi-objective optimization problem, this solution is said to be a non-dominated solution.

Let 푀 be a feasible assignment matrix for a section of path. A feasible assignment matrix 푀∗for a multi objective optimization problem, say optimize 풙 = (푥 , 푥 , … , 푥 ) is a non dominated (Pareto) solution if there is none feasible assignment matrix 푀 such that 푥 (푀) improves 푥 (푀∗), 푘 ∈ {1,2, . . . ,푝}, and ∃푚 ∈ {1,2, . . . ,푝},푥 (푀) ≠푥 (푀∗). For a p objectives assignment optimization problem, the number of feasible solutions may grow exponentially with the number of users [5].

To avoid the difficulty of the multiplicity of Pareto optimal solutions, a utility function is proposed. In the evolutionary approach, the utility function is taken as the fitness function to be maximized/minimized and it is a function about the variables with which the results evolve. And it is possible and necessary to make homogeneous the different variables that will influence the user’s decision. For each modality of one section of the route, the variables like travel cost, travel time etc. are referred as the attributes of the route.

It is usually assumed that the fitness is a linear function of the attributes of the route [5], that is 푓 푥 , 푥 , . . . , 푥 = ∑ 훽 푥 , where 푥 is the value of the 푘th attribute and 훽 is the coefficient of 푥 . It exists other similar forms of fitness function that will be discussed later.

Note that, for one segment 푆(푡) , which is represented by an arc in the graph, its attributes like cost and time will be expressed in the fitness function. Apart from these involved attributes, the issues about the homogenization and the ratio between the coefficients should be considered.

When the EAs are applied to the multi-objective optimization problem, a fitness value of each solution should be evaluated. The fitness function of the solution 풙 can be defined by the following weighted sum of the 푛 objectives:

푓(풙) = 휔 푓 (풙) +휔 푓 (풙) +⋯+ 휔 푓 (풙) (2)

where 휔 ,휔 , . . . ,휔 are nonnegative weights for the 푛 objectives, which satisfy the following conditions:

∀푖 ∈ {1,2, . . . ,푛},휔 ≥ 0and 휔 = 1

Each objective can be any function of the solution 풙. In our case we take 푓 (푥) = 휈 /푥 with the search direction weight vector 풘풃 = (휔 ,휔 , . . . ,휔 ).

Then the fitness function is transformed to the following form:

푓(풙) = 휔 휈 /푥 + 휔 휈 /푥 +··· +휔 휈 /푥 (3)



where 휈 are the constants that make the different objectives homogeneous. Here, we can take 휈 = 푓 (풙∗) where 푓 (풙∗) is the optimal solution related to the objective function 푓 .

The coefficients are used for weighting the different parameters in a way how they are perceived by the travelers. Without loss of the generality, we assume that the fitness function is to be minimized to get the optimal solution. In other words, the satisfaction of the traveller increases as the value of attribute increases.

The steps involved in this evolutionary algorithm for the vehicles assignment problem are as follows:

1. Generate an initial population of 푁 randomly constructed solutions. Each of the initial solutions is generated in a random way. Since initial solutions may violate the vehicles’ capacity constrains, they may be infeasible.

2. A fitness function is defined with the method of aggregation. 3. Decode the solution structure. The assignment of the travelers to the vehicles in one route is

represented in form of matrix. The fitness of each solution is calculated according to the fitness function defined before. Apart from the aggregation technique for the variables, it must take into account the degree of infeasibility that is a frequent approach in EAs. In evolutionary computation most of the constraint-handling methods are based on the concept of penalty functions [14]. In this case, when the assignment solution exceeds the number of available vehicles or vehicles’ capacity, a penalty will be imposed on the fitness.

4. Select two parent matrixes in the parent generation for reproduction. A roulette wheel selection method is adopted. On the roulette wheel each individual is represented by a space that proportionally corresponds to its fitness [15]. In a roulette wheel selection, the fittest individuals have a greater chance of survival that weaker ones. The fitter individuals will tend to have a better probability of survival and will go forward to from the mating pool for the next generation. To preserve the diversity, the weaker ones are not without chance because they may have genetic coding that proves useful to future generations.

5. Generate a child solution by applying firstly a crossover operator (combination) to the selected parents. The one-point crossover is chosen for this procedure in which a point for the crossover p is selected randomly and the offspring solution will make up of the first part from the first parent and the second part taken from the other parent, or vice versa with equal probability. Globally, the crossover operator occurs with a certain probability. After the crossover operator, a mutation operation follows. In this procedure, it involves a mutation in a randomly selected gene, with a certain probability. Then the procedure of the correction of the mutation is executed to verify if the generated children are legal, which means that the constrains are satisfied. If not, the children chromosomes will be slightly modified.

6. Replace the individuals in the population by the child solutions, and then the offspring generation is obtained.

7. Repeat steps 4-6 until the best solution cannot be improved any more.

After selection step has been carried out, another population that is referred as intermediate



www.elvedit.com 92

population in [15] is completely constructed, and then the crossover can occur. This crossover operator is applied to randomly paired chromosomes denoted 푃 . These chromosomes are recombined to two new matrixes with a probability 푃 .

Consider the following chromosomes Parent 1 and Parent 2:

Parent 1 Parent 2 푅(푥, 푦) 푉 푉 푉 푅(푥, 푦) 푉 푉 푉 푃 0 ∗ 1 푃 0 1 ∗ 푃 1 ∗ 0 푃 1 ∗ 0 푃 ∗ ퟏ ∗ 푃 ퟏ ∗ ∗ 푃 0 ∗ 1 푃 0 1 ∗

These two matrixes represent two possible solutions to the affectation problem. Then using a single randomly chosen crossover point, the one-point operator takes place. Two offsprings are obtained in swapping the fragments between the two parents.

Offspring 1 Offspring 2 푅(푥, 푦) 푉 푉 푉 푅(푥, 푦) 푉 푉 푉 푃 0 ∗ 1 푃 0 1 ∗ 푃 1 ∗ 0 푃 1 ∗ 0 푃 ퟏ ∗ ∗ 푃 ∗ ퟏ ∗ 푃 0 ∗ 1 푃 0 1 ∗

The mutation operator follows the recombination procedure. For each bit in the population, mutate with a certain low probability Pm. The following example will illustrate how this operator works.

Matrix before mutation

Matrix after mutation Matrix after correction of mutation

푅(푥, 푦) 푉 푉 푉 푅(푥, 푦) 푉 푉 푉 푅(푥,푦) 푉 푉 푉 푃 0 ∗ 1 푃 0 ∗ 1 푃 0 ∗ 1 푃 1 ∗ 0 푃 1 ∗ 0 푃 1 ∗ 0 푃 ∗ ퟏ ∗ 푃 ∗ ∗ ퟏ 푃 ퟏ ∗ ∗ 푃 0 ∗ 1 푃 0 ∗ 1 푃 0 ∗ 1

After the mutation, some chromosome may become inappropriate for that the number of the passengers may exceed the capacity of the vehicle. In this case, the relevant affectation should be modified with the vehicle’s volume and the user’s preference.

D. An Alternative Optimization Approach

The time of execution depends on the complexity of the algorithm and the size of the problem. To obtain a better performance of calculation, an optimization approach for this problem is proposed in case that the size of the requests at the same time is very limited and there is no need to launch EAs. At a time the requests are launched, the system will choose the appropriate algorithm.

The local search algorithm is one algorithm frequently used in the filed of scheduling. The



FIGURE III: THE GRAPH OF THE DEMANDED ITINERARIES

FIGURE IV: THE DECOMPOSITION ACCORDING TO THE AREAS AND THE MODES

preferences of the users and the availability of the transportation tools are the only elements to consider. Briefly, the principle rule is “first arrived, first served”. The users are ranked in a certain order when they launch their itinerary queries. On the basis of this ranking the allocation is performed. With this algorithm, the vehicles will be allocated to the users according to their choices. The first choice is satisfied if the relevant vehicles are available; if not, the second choice will be considered, and so forth. The demanders’ preferences will be satisfied as much as possible.

E. Switching between algorithms

For the sake of performance and response time, the algorithms will be chosen automatically. The criterion to follow is the size of the problem and the availability of the resource to allocate. In the default case, the evolutionary algorithm is used; otherwise, when the quantity of requests is too limited to take full advantage of EAs or even the resource is too limited, the alternative approach will be chosen.

V. SIMULATION RESULTS In order to illustrate how the system works and to evaluate the approach proposed in this

paper, a simulation example of requests will be presented as follows. In the simulations, here are the parameters for EAs:

-Population size: 푁 = 100 -Number of generations: 푁 = 100 -Mutation probability: 푃 = 0.05 -Crossover probability: 푃 = 0.80

As showed in Figure 3, there are 6 itineraries requests for the correspondent time window and we have gotten the shortest itinerary for each request. Here are the requests, 퐼(푡 ) = {퐼(푡 ) , 퐼(푡 ) , 퐼(푡 ) , 퐼(푡 ) , 퐼(푡 ) , 퐼(푡 ) }. 퐼(푡 ) = 퐴 → 푂[푡 , 푡 ],withoutpreference; 퐼(푡 ) = 퐵 → 퐾[푡 , 푡 ],nofree-usecar; 퐼(푡 ) = 퐶 → 퐽 푡 , 푡 ,carpooling irst; 퐼(푡 ) = 퐷 → 푀[푡 , 푡 ],free-usecar irst; 퐼(푡 ) = 퐸 → 퐿[푡 , 푡 ],nocarpooling; 퐼(푡 ) = 퐹 → 푁[푡 , 푡 ],nopublictransport.

After having identified the requests, we decompose the itineraries according to the geographical elements and the transportation modes. Thus, there are three areas and three general types of transportation in the decomposition, as showed in Figure 4.



www.elvedit.com 94

In total, five transportation service providers are concerned across the transportation network, we name 퐶 for public transportation in area I, 퐶 for public transportation in area II, 퐶 for public transportation in area III, 퐶 for carpooling service and 퐶 for free-use car service. Therefore the itineraries can be described in the following form: where 푚 represents public transport, 푚 and 푚 represent carpooling and free-use car, respectively. 퐼(푡 ) = (퐴,퐺) , , (퐺,퐻) , , , (퐻,푂) , 퐼(푡 ) = (퐵,퐺) , (퐺,퐻) , , , (퐻,퐾) , , , , 퐼(푡 ) = (퐶,퐺) , (퐺,퐻) , , , (퐻, 퐽) , , 퐼(푡 ) = (퐷,퐺) , , (퐺,퐻) , , , (퐻,퐾) , , 퐼(푡 ) = (퐸,퐺) , , (퐺,퐻) , , , (퐻, 퐿) , 퐼(푡 ) = {(퐹, 퐼) , (퐼,퐻) , (퐻,푁) , }

There are several route segments, (퐴,퐺), (퐷,퐺), (퐸,퐺), (퐹,퐺), (퐺,퐻), (퐻, 퐽), (퐻,퐾), (퐻, 퐿), (퐻,퐾), etc. which are assured by more than one transportation service. The segments 퐺퐻 are shared by several itineraries; thus, the assignment process will be launched to allocate the resource. As is showed in the graph, it’s in a distributed environment. The transportation service is provided by several operators. After the decomposition of the itineraries, the assignment will occur for each individual segment. In the following paragraphs, two examples will be presented to illustrate how the assignment works. For the common segment (퐺,퐻) for the requests 퐼(푡 ) , 퐼(푡 ) , 퐼(푡 ) , 퐼(푡 ) , 퐼(푡 ) , the EAs will be applied for the assignment. We note the relevant time window is 푇 . For the segment, (퐻,푁) for the request 퐼(푡 ) , the local search algorithm takes in charge of the assignment.

In this simulation, three parameters will be taken into account as optimization criterions. The travel cost, the traveling time and the comfort satisfaction for each segment are considered. As mentioned in section IV for the fitness function in EAs, the parameters will be homogenized so that they can be added. Also the weight that the parameter presents in the fitness function is a multiplication coefficient and is predefined. Take one route from G to H during the specified time window T as example. The characteristics for each proposed transportation tools are shown in Table 1.

TABLE I: CHARACTERISTICS OF DIFFERENT TRANSPORTATION TOOLS FROM G TO H DURING T 푅 (퐷, 퐿) Train Carpooling Free-use car

Cost (euros) 35 18 30 Time (mins) 70 120 90

Comfort 2 4 1

Here, we introduce a characteristic of the transportation tools: the comfort. It is the index of personal feeling of the user when using this method of transport. It varies from “1” to “5”, meanwhile “1” means the most comfortable and “5” means the least comfortable.

There are 5 transport demands for (퐺,퐻) are 2 carpooling places and 1 free-use car. According to the definition in section IV and the data in Table 1, the fitness function is:

푓(풙) = 휔 휐 /푥 + 휔 휐 /푥 + 휔 휐 /푥 (4)

where 휐 = 18, 휐 = 70, 휐 = 1, 푥 is the cost, 푥 is the time, 푥 is the comfort level. For the weight vector of the fitness function, we have 풘풃 = (0.5, 0.3, 0.2). Thus, the weight parameters for the fitness function are as follows: travel cost 휔 = 0.5, travel time 휔 = 0.3 and comfort condition 휔 = 0.2. The fitness function becomes:



푓(풙) = 9/푥 + 21/푥 + 0.2/푥 (5)

After the assignment with the EAs, a possible assignment matrix is obtained as the Table. It’s possible that it exists more than one feasible solution; in this case our future work for the coalition of the segments will concentrate on choosing the best one. After the assignment process for each route, the following result for each request is gotten for each request. 퐼(푡 ) = (퐴,퐺) , , (퐺,퐻) , (퐻,푂) , 퐼(푡 ) = (퐵,퐺) , (퐺,퐻) , (퐻,퐾) , , , 퐼(푡 ) = (퐶,퐺) , (퐺,퐻) , (퐻, 퐽) , 퐼(푡 ) = (퐷,퐺) , (퐺,퐻) , (퐻,퐾) , 퐼(푡 ) = (퐸,퐺) , (퐺,퐻) , (퐻, 퐿) , 퐼(푡 ) = {(퐹, 퐼) , (퐼,퐻) , (퐻,푁) }

Another simulation example is proposed in the local transportation network of Lille. The transportation flux is sometimes not avoidable. A strong demand of transport from near cities to Lille is also a problem. For the transportation between these cities, as in Figure 5, apart from private cars, there are regional train (Ter), free-use car and carpooling. At the same time, the carpooling can take passengers to the destination (Grand Stade), the free-use cars should be returned to the service center. The cost and the time needed for these transportation methods are as the following Table 2.

TABLE II: CHARACTERISTICS OF DIFFERENT TRANSPORTATION TOOLS FROM DUNKERQUE TO LILLE IN T

푅 (퐷, 퐿) Train Carpooling Free-use car Cost (euros) 15 10 25 Time (mins) 75 65 55

Comfort 2 4 1 According to the definition in section IV and the data in Table 2 and the weight vector of the

fitness function, we have 풘풃 = (0.5, 0.3, 0.2). The fitness function becomes:

푓(풙) = 5/푥 + 19.5/푥 + 0.2/푥 (6)

The number of requests varies significantly. The response time of the algorithm changes along with the quantity of requests. Figure 6 is the relation between the response time and request quantity. The response time is approximately linear with the request quantity. The proposed encoding method and EAs can cope with this type of problem without too long response time even big quantity of requests.

FIGURE VI: THE RESPONSE TIME AND THE REQUEST QUANTITY

FIGURE V: DEPARTURE AND ARRIVAL SPOTS OF THE REQUESTS



www.elvedit.com 96

The following scenarios will show the respect of the travelers’ preference and the rate of their satisfaction. The application of the EAs gives the following results:

(1) In case of sufficient resource of carpooling and free-use car (we consider the places in train are unlimited), the EAs will satisfy the travelers’ choice and optimize the fitness function defined above. The fitness function is maximized with respect to the condition of availability of the relevant vehicles. The preference is given as Table 3, and we get Table 4 as the assignment matrix. The evolution of the fitness function is as the Figure 7. The rate of satisfaction of all the travelers is 100%.

(2) In case of insufficient resource of carpooling or car sharing, the applied algorithm will firstly meet the travelers’ demands that are restrictive to only one method of transportation. Then the fitness function is maximized with respect to the condition of availability of the relevant vehicles. There are 30 service requests; meanwhile 5 carpooling places and 5 free-use cars are available. The users’ preference is given as Table 5. Thus, we get Table 6 as the assignment result. The evolution of the fitness function is as the Figure 8. With the condition of the availability of the vehicles, the specific demand is met. Two requests with their first choices for carpooling and another two requests for free-use cars are allocated with train, all the rest satisfied for their first preferences.

(3) In case of insufficient resource of both carpooling and car sharing, the local search algorithm will applied. In the limit of the availability of the vehicles, the preference of the requests will be met as much as possible. There are 30 service requests and 5 carpooling places and 5 free-use cars available. The preference is given as Table 7, and we get Table 8 as the assignment matrix respectively. With the condition of the availability of the vehicles, the specific demand is met. Two requests with their first choices for carpooling and another two requests for free-use cars are allocated with train, all the rest are satisfied for their first preferences.

As the local search algorithm only charges the factors of preference and availability, no complicated combinatory will occur. This is reliable and feasible for little quantity of requests and insufficient vehicle resource.

FIGURE VII: THE EVOLUTION OF THE FITNESS FUNCTION

FIGURE VIII: THE EVOLUTION OF THE FITNESS FUNCTION



TABLE III: PREFERENCE OF THE REQUESTS

TABLE V: PREFERENCE OF THE REQUESTS

TABLE VI: ASSIGNMENT

RESULT

TABLE IV: ASSIGNMENT

RESULT

TABLE VII: PREFERENCE OF THE REQUESTS

TABLE VIII: ASSIGNMENT

RESULT

VI. CONCLUSIONS AND FUTURE WORKS In this paper, an algorithm for the assignment of vehicles in a distributed co-modal transport

information system is presented. The preference and the availability are both considered while the optimization. With the evolutionary algorithm and the local search algorithm, the system switches between them in favor of the response time. The part of simulation shows how the system works.

In the future work, the coalition of the segments will be studied. The combination of different routes represented by the Route Agent will be studied. This coalition is dedicated to the formalization of an optimized itinerary solution for the travelers. Especially, a protocol for the communication between the agents, which represent the segments, will be established. After the coalition procedure, a complete journey will be returned with the employed transportation

푅 푉 푉 푉푇 1 0 0 푇 1 0 0 푇 0 1 0 푇 0 1 0 푇 0 1 0 푇 0 0 1 푇 0 0 1 푇 0 0 1 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 ∗ 1 ∗ 푇 ∗ ∗ 1 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗

푅 푉 푉 푉 푇 1 0 0 푇 1 0 0 푇 0 1 0 푇 0 1 0 푇 0 1 0 푇 0 0 1 푇 0 0 1 푇 0 0 1 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 ∗ 1 ∗ 푇 ∗ ∗ 1 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ∗ ퟏ ∗ 푇 ퟏ ∗ ∗ 푇 ∗ ∗ ퟏ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗

푅 푉 푉 푉푇 1 0 0 푇 1 0 0 푇 0 1 0 푇 0 1 0 푇 0 1 0 푇 0 0 1 푇 0 0 1 푇 0 0 1 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 ∗ 1 ∗ 푇 ∗ 1 ∗ 푇 ∗ 1 ∗ 푇 ∗ 1 ∗ 푇 ∗ ∗ 1 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗

푅 푉 푉 푉 푇 1 0 0 푇 1 0 0 푇 0 1 0 푇 0 1 0 푇 0 1 0 푇 0 0 1 푇 0 0 1 푇 0 0 1 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 ∗ 1 ∗ 푇 ∗ 1 ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ∗ ∗ 1 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ∗ ∗ ퟏ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗

푅 푉 푉 푉푇 1 0 0 푇 1 0 0 푇 0 1 0 푇 0 1 0 푇 0 1 0 푇 0 0 1 푇 0 0 1 푇 0 0 1 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 ∗ 1 ∗ 푇 ∗ 1 ∗ 푇 ∗ 1 ∗ 푇 ∗ 1 ∗ 푇 ∗ ∗ 1 푇 ∗ ∗ 1 푇 ∗ ∗ 1 푇 ∗ ∗ 1 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗ 푇 ∗ ∗ ∗

푅 푉 푉 푉 푇 1 0 0 푇 1 0 0 푇 0 1 0 푇 0 1 0 푇 0 1 0 푇 0 0 1 푇 0 0 1 푇 0 0 1 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 1 ∗ ∗ 푇 ∗ 1 ∗ 푇 ∗ 1 ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ∗ ∗ 1 푇 ∗ ∗ 1 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗ 푇 ퟏ ∗ ∗



www.elvedit.com 98

service.

REFERENCES [1] European Commission. European transport policy for 2010: time to decide. White Paper.

[2] Florez-Mendez, R. A. (1999). Toward a Standardization of Multi-Agent System Frameworks. In ACM Crossroads Student Magazine, Canada.

[3] Green, S., Hurst L., Nangle, B., Cunningham, P., Somers, F., Evans, R. (1997). Software agents. A review, TSC-CS-1997-06, Trinity Collège Dublin, roadcom Eireann research Ltd.

[4] Kamoun, M. A., Uster, G., Hammadi, S. (2005). An agent-based cooperative information system for multi-modal travelers route computation. Systems, Man and Cybernetics, 2005 IEEE International Conference on, 21,162-67.

[5] Modesti, P., & Sciomachen, A. (1998). A utility measure for finding multi-objective shortest paths in urban multimodal transportation networks. European Journal of Operation Research, 111(3), 495-508.

[6] Wang, J., & Kaempke, T. (2004). Shortest route computation in distributed systems. Computers & Operations Research, 31(10), 1621-33.

[7] Wang, J., & Kaempke, T. (2006). The fastest itinerary in time-dependent decentralized travel information systems. Journal of Combinatorial Optimization, 12(3), 167-185.

[8] Sghaier, M. (2011). Combinaison des techniques d’optimisation et de l’intelligence artificielle distribuée pour la mise en place d’un système de covoiturage dynamique (Doctoral dissertation, Ecole Centrale de Lille). Retrieved from http://www.theses.fr/2011ECLI0021

[9] Ayed, H., Galvez-Fernandez, C., Habbas, Z., Khadraoui, D. (2011). Solving time-dependent multimodal transport problems using a transfer graph model. Computer & Industrial Engineering. 61(2), 391-401

[10] Jeribi, K., Zgaya, H., Zoghlami, N., Hammadi, S. (2011). Distributed architecture for a co-modal transport system. In Systems, Man, and Cybernetics, 2011 IEEE International Conference on 2797-2802.

[11] Jeribi, K., Mejri, H., Zgaya, H., Hammadi, S. (2011). Distributed graphs for solving co-modal transport problems. Intelligent Transportation Systems (ITSC), 14th International IEEE Conference on, 1150-55.

[12] Ishibuchi, H., & Murata T. (1998). A multi-objective genetic local search algorithm and its application to flowshop scheduling. IEEE Transactions on System, Man and Cybernetics, 28(3), 392-403.

[13] Ulungu, E.L., & Teghem, J. (1994). Multi-objective combinatorial optimization problem: a survey. Journal of multi-criteria Decision Analysis, 3, 83-101.

[14] Michalewicz, Z., & Schoenauer, M. (1996). Evolutionary algorithms for constrained parameter optimization problems. Evolutionary Computation, 4(1), 1-32.

[15] Whitley, D. (1994). A genetic algorithm tutorial. Statistics and Computing, 4, 65-85.




Zhanjun Wang is actually a Ph.D student within LAGIS at the Ecole Centrale de Lille (French “Grande Ecole”). His current research is on the distributed optimization, transport information system and distributed artificial intelligence. He is born in Kaifeng (China) in 1987. He received the General Engineer degree from Ecole Centrale de Pékin and the Master’s degree in electrical and electronic science from Beihang University

(China) in 2012.

Dr. Khaled MESGHOUNI was born in Constantine, Algeria in 1968, he obtained the Diploma of "Ingénieur at University of Constantine (Algeria), Master degrees in Electrical engineering at Ecole Centrale de Lyon (France). He obtained Ph.D. in Automatic Control and Computer engineering of the University of Science and Technology of Lille (France) in 1999 and the HdR “Accreditation to supervise Ph.D thesis” in the same University in 2007. Actually he is an Associate Professor, his current research interests focus on the area of aartificial intelligent, production planning, manufacturing systems, advanced

mobility and transport engineering.

Mr. Slim Hammadi is a full Professor of production planning and control at the Ecole Centrale de Lille (French « Grande Ecole »). Born in Gafsa (Tunisia) in 1962, he has obtained by 1988 the Master degree in Computer science from the University of Lille (France). Pr Hammadi obtained a P.h.D degree in job-shop scheduling and control in 1991 at Ecole Centrale de Lille. He is a senior member of IEEE/SMC and has served as a referee for numerous journals including the IEEE Transactions on SMC. Pr. S. Hammadi was Co-Organizer of a Symposium (IMS) of the IMACS/IEEE SMC Multi conference

CESA‟98 held in Hammamet (Tunisia) in April 1998. He has organized several invited sessions in different SMC conferences where he was session chairman. He was chairman of the International congress on “Logistic and Transport” LT‟2004, MHOSI‟2005, LT‟2006 and LT‟2007. His teaching and research interests focus on the areas of production control, production planning, computer science, discrete and dynamic programming and computer integrated manufacturing.


Online Brand Experience Creation Process Model: Theoretical Insights

Authors



Mindaugas Kiskis Institute of Digital Technologies, Faculty of Social Technologies, Mykolas Romeris University


Virginija Jurkute Institute of Digital Technologies, Faculty of Social Technologies, Mykolas Romeris University


Abstract

Many brands are turning digital due to the changing market requirements and consumer demands. In order to digitize the brand, it is not enough just to move the brand to the electronic environment. Marketing plans and other brand activities shall be revised and adopted to the electronic environment. The focal point for the digital transformation of the brand is the online brand experience. It is increasingly recognized as a vital tool for the success of the brand. The impact of brand experience on the consumer trust and loyalty is empirically proven and explained in existing research, however the process of the online brand experience building is not well understood and in practice based on trial-and-error rather than research framework.This paper studies conceptual issues of the online brand building. Online brand experience concept is examined in order to set the framework for the online brand creation model. The study reveals that online brand experience may be based on the traditional brand experience models, that is - consumer’s perceptions and responses to brand evoked stimuli. This definition is assumed for further analysis of the online brand creation process. Comparative analysis of existing brand experience creation models allows identification of the main building blocks and creation steps for the online brand experience. The paper concludes that online brand experience creation is based on the adaptation of the traditional marketing models (“4P” marketing elements) to the specifics of the online environment and processes. The modified model nicknamed 3PoP is proposed. The 3PoP model embraces the 3 traditional P’s - product, place, people, filtered through the online process as the core of the online brand creation. The 3PoP model enables further research and management applications leading to the holistic online brand experience.

Key Words Online brand, brand experience, consumer experience, marketing mix, online brand experience building blocks.

Online Brand Experience Creation Process Model: Theoretical Insights Tadas Limba, Mindaugas Kiskis, Virginija Jurkute


I. INTRODUCTION Significance of brand is doubtless in marketing context, especially for the online business [8, 15,

9]. Its essence is fulfilled through such brand constructs as brand attitudes, brand involvement, brand attachment, customer delight, brand personality and brand experience [5, 48]. As more customers are getting used to shopping online, brand managers are shifting their focus to digitization of the brand and the attributes, which could increase the brand attractiveness and satisfy their customer’s needs. The most significant construct for the brand attractiveness traditionally is brand experience [25, 16]. For the online commerce this means that customer’s offline store experience has to be transferred to the online environment, thus transforming costumers into the online brand consumers. Consumption is very much related to the experience. Each brand brings it to their consumers through interaction or touch points. Online businesses are highly restricted with the interaction points and the area where they can use them to create consumer experience. Mainly it can be done only through the screen, which is seen by the consumer. The web sites are the first point of consumer and brand online interaction. They are the primary tools engaging consumers. By implementing them in the correct manner and based on the effective principles, online brand experience can be much increased, and consumer’s trust [20, 21] and loyalty [21, 22] ensured. De Chernatony and Horn see experience as essential condition of brand existence defining brand as “<…a cluster of functional and emotional values which promise a particular experience>” [18, p. 1100]. In A. T. Kearney global management consultancy and executive search firm white paper it is highlighted that “<…there is no better way to create customer loyalty and enthusiasm than through a positive … experience.>” [29, p. 1]. Brand experience is met in all touch points of the brand and its consumers, starting from searching (how it is easy or difficult to find it), ending with purchasing and services thereafter [14]. So it is vital to ensure brand experience which brings positive consumers attitude towards the brand. It can assure greater chance of success for the brand owner, but also in case of failure, it can cause a huge damage [42]. It can impact brand and consumers equity growth [3, 29] and can be used as the main differentiator from the competitors [4, 7]. These advantages of brand experience have already attracted a lot of the marketer’s attention [5]. Recent studies have presented its significance to the development of marketing strategies [9, 49].

Scientific issue. Despite the fact that brand experience is becoming one of the most discussed topic in studies of marketing assets to ensure successful business in online environment, there is a lack of commonalities in the existing studies, which describe how to create and manage it [23, 48]. Several papers draw attention to the creation process of the brand experience and provide managerial suggestions about the factors, which should be fulfilled to ensure great brand experience for its consumers. But all of the given models provide different view of the brand experience factors and guidance’s for their implementation. Additionally there is a lack of empirical research of online brand experience creation and managing process. Quite often brand experience is used interchangeably with consumer experience or web experience [13, 29, 42]. Thus, a question of online brand experience concept is proposed along with the analysis of whether the traditional brick and mortar commerce brand experience concept can be used for the online brand experience definition. The answer to this question is necessary in order to have clear understanding of online brand experience, which provides the background for the creation



www.elvedit.com 102

and management process of the online brand experience. Finally, investigation of the building factors of the online brand experience defines the fundamental building blocks vital for the brand creation process online.

Object of the research – key features and the building blocks of the online brand experience.

Purpose – is to propose the online brand experience creation process model, which shall encompass essential positive brand experience building blocks and ensures consumer satisfaction, trust and loyalty.

Objectives: 1. To analyze brand experience definitions and conceptualize online brand experience; 2. To conduct comparative analysis of brand experience creation models, and to identify the

most important factors for successful online brand experience creation; 3. To propose the online brand experience creation process model and to identify the main

building blocks thereof.

Methodology – the paper relies on the theoretical-systematic, conceptual comparative and phenomenological analysis as well as meta-analysis of the previous studies. A method of dynamic modeling is also applied.

Practical significance – the online brand experience creation process model proposed in the paper may serve for the business foundation of the comprehensive approach to online brand experience, as well as set the planning guidelines and responsibilities.

II. DEFINING THE ONLINE BRAND EXPERIENCE Brand experience is becoming synonym of a brand [43]. Sometimes it is used interchangeable

with brand familiarity [20] or mixed with other brand constructs such as brand attitudes, brand involvement, brand attachment, or brand personality [5]. This misunderstanding can be derived from insufficiently defined brand experience concept used in managerial and scientific literature. It is essential to clearly define the online brand experience concept in order to design the framework for the brand experience creation process. Not many studies on online brand experience exist to date. Brand experience in the traditional brick and mortar environment is commonly investigated as the basis for the online brand experience. The analysis of several brand experience definitions are given below in chronological order (see Table 1).

It is noticeable that conceptualization of the brand experience is closely related to the customer experience. But as Kapferer characterizes online purchaser as consumer, not a customer, so this concept will be used in further study [28]. Consumer concept is more suitable to describe the gaining ways and outcomes of the online brand experience. Experience in most cases is viewed through a person who is engaging with a brand, so consumer experience can be considered as concurrent with brand experience, and used interchangeable. As Wright et all highlights “<people do not simply engage in experiences as ready-made, they actively construct them through a process of sense making…>” and reflect sensually, emotionally and behaviorally [48].



This is already represented in the definitions given above, which encompass behavioral, sensational, cognitional and emotional aspects. These aspects are most related to consumer perceptions and responses towards brand acquired through the time and shaped by marketing activities, which Klaus and Maklan calls holistic offer of a brand [30].

TABLE I: BRAND EXPERIENCE DEFINITIONS Author Definition

Smith and Wheeler, 2002, p. 6

Brand experience is: intentional in delivering a customer experience to support the brand; consistent in delivering that experience over time and location; differentiated from competing brands; valuable in offering a customer proposition which meets target customer needs.

Ha and Helen Perks, 2005, p. 440 Brand experience is a coalescing of symbolic meaning with allied behavior, thoughts and feelings that occur during the consumption of the service/product.

Alloza, 2008, p.373

Brand experience is the perception of the consumers, at every moment of contact they have with the brand, whether it is in the brand images projected in advertising, during the first personal contact, or the level of quality concerning the personal treatment they receive.

Brakus et all, 2009, p. 53

Brand experience is subjective, internal consumer responses (sensations, feelings, and cognitions) and behavioral responses evoked by brand-related stimuli that are part of a brand’s design and identity, packaging, communications, and environments.

Iglesias and Singh, 2011, p. 571 Brand experience is a takeaway impression (Carbone and Haeckel, 1994) that is formed in the mind of the consumers as a result of the encounter with the holistic offer of a brand (Klaus and Maklan, 2007).

Sahin et all, 2011, p. 1290 Brand experience is not an emotional relationship concept. Experiences are sensations, feelings, cognitions, and behavioral responses evoked by brand related stimuli.

Brand experience is often mixed with other brand elements mentioned above because of its inputs and outcomes. Nevertheless it should be admitted as distinct construct, especially from the perspective of brand personality, which is defined as brand associated human characteristics [1], originating from the brand involvement, because <…brands that consumers are highly involved with are not necessarily brands that evoke the strongest experiences> [5, p. 53]; from brand attachment, which encompasses strong consumer’s affection to brand and refers to emotional relationship concept [5, 39]; and from brand familiarity, which is more the outcome of the brand experience [20].

Examination of all given definitions above, suggests common features for all of them. Alliance



www.elvedit.com 104

of the brand experience with the consumer’s behavioral, emotional and sensational responses, which are evoked by the brand related stimuli, is the joining link. It comes via the product experience occurring during the search, purchase or interaction with the product/service in other ways (i.e., through advertisement or when the product is presented virtually) process, shopping, service and consumption experiences, which are the parts of overall brand experience [5]. These experiences raise internal and external consumer’s reactions. Overall, the above analysis enables to conceptualize brand experience as the consumer’s perceptions and responses to brand evoked stimuli.

Switching focus to the business in electronic environment the concept of the web experience is commonly encountered. Constantinides explains web experience as follows – „<…web experience: a combination of online functionality, information, emotions, cues, stimuli and products/ services, in other words a complex mix of elements going beyond the 4Ps of the traditional marketing mix>” [13, p. 112]. Operating in such environment the term of web experience can be used interchangeably with the online brand experience, because this is the primary tool for the businesses to ensure consumers interaction with a brand. The website is the first touch point which can be seen and tried by the consumer, therefore this is the place for the online brand experience creation. Combination of functionality, information, products, services and other stimuli impacts consumer’s responses and provided information, emotions, cues are the sources of perception. Constantinides web experience concept definition is closely related with traditional brand experience concept, expressing consumer’s perceptions and responses to brand evoked stimuli, but it might focus marketers too much on the web site technical solutions, leaving the strategic decisions about brand experience creation in this environment [13].

Ha and Perks also emphasize web site as a primary touch point [20]. The experience gained through this tool they identify as the website brand experience and define it as follows: „<...a consumer's positive navigations (i. e. using web-based communities and participating in events) and perceptions (i. e. the attractiveness of cookies, variety and uniqueness of visual displays and value for money) with a specific website>” [20 p. 440]. Thus, the main task for the web site is to create positive experience to consumers, which reassures that consumers return to a particular web site as to the favorite one. Analyzing this definition, the necessary actions, such as, positive navigation of the consumers, to be ensured by the marketers, is mentioned. This definition provides the cues to the outcomes of company’s insights and strategy, focusing not only on technical web site decisions, but also considering, what could ensure consumers satisfaction in different brand touch points. That, as Ha and Perks highlight, is the preliminary condition for consumers to participate in e-commerce [20].

Kahn characterizes experience as a “<…result of encountering, undergoing or living through situations” [26, p. 14]. More explanations of this concept can be found in Jack Morton Worldwide global brand experience agency “Best Experience Brands” study, which provides brand experience term definition as “<…it is referring to any of the interactions you have with either the specific company or its products or services. This can include your own personal use of the product or brand, your conversations or interactions with employees or people who represent the



brand or anything you learn from that brand’s marketing, word-of-mouth, recommendations from your friends, colleagues or social network.>” [24, p. 6]. This comprehensive definition is very close to the online brand experience concept used by A. T. Kearney, which emphasizes that this experience encompasses all consumers’ interaction points with online business [29]. These brands “touch points” can be defined as brand related stimuli. According Jack Morton Worldwide study the main important stimuli, or in other words drivers of online brand experience are: efficient purchasing experience; products and services that meet the consumer needs; easiness of buying services, whenever and wherever consumer is shopping; understanding consumer needs; exceeding their expectations; and initial impression the brand makes on consumers [24].

Summarizing the above analysis, it may be concluded that some of the definitions are more related to the stimuli, which drive the brand experience, rather than explaining what online brand experience is. The other definitions are focused more on the technical side of the online brand experience, such as web site usability, design or information relevance, rather than strategic approach to the online brand experience creation. The commonalities for all of them are brand evoked stimuli and consumers' perceptions and responses to these drivers. Thus, traditional notion of the brand experience is still useful for the definition of the online brand experience concept. The difference is only in its implementation process [29], because of the distinct brand and consumer touch points. In subsequent study this online brand experience definition – consumers perceptions and responses to brand evoked stimuli, will be used, focusing to the process and drivers which could create positive experience, ensuring brand familiarity, satisfaction, brand trust and set the conditions for consumers high involvement in participation in e-commerce.

III. COMPARATIVE ANALYSIS OF BRAND EXPERIENCE CREATION MODELS Internet had expanded the space for the evolution of business. Also, it set specific additional

rules for the making business. Although the online consumers might be the same as in the traditional environment, their online behavior differs. Kotler highlights that “<internet consumers have around-the-clock access to varied information sources, making them better informed and more discerning shoppers>” [31, p.235]. The consumers are demanding more and more from the online brands [15]. They want not only to get the information about the brand, but also experience it while visiting the web site, to talk about what they had experienced through the different stages of getting familiar with a brand, share that experience in online communities, which they trust more than brand owners [19, 20, 45]. Thus, brand experience here is becoming one of the most influential marketing paths. It is empirically proved that brand experience is the primary factor to create consumers trust and loyalty [5]. It is recognized as a vital driver to the business success online [39]. The companies such as Amazon acknowledged this early during their evolution and thus captured the leading positions among online retailers by providing most experiential brand. New generation businesses, like Amazon, realized that the process of how is“<…more critical than what is being sold>” [29, p. 3]. The question of how to provide positive brand experience for the consumers was subsequently raised in the online marketing world. The correct answer to this question is that positive brand experience is the tool to assure high results of business performance not only by increasing conversion into sales rates



www.elvedit.com 106

[10, 38], but also by enlarging the volume of loyal consumers, who broaden brand existence boundaries and help to occupy top brand positions. Positive brand experience provides premises for brand familiarity, satisfaction and influences further purchases [36]. While purchasing consumers expect not only rational benefits, but also memorable brand experience [34]. To meet their expectations companies need to clarify how to provide such experience for the consumers of their brands. Experience might be gained in physical, audible and visual ways [46], therefore analysis of the use of these means is necessary to have all-inclusive picture of online brand experience creation process. Several creation models are presented and analyzed below. They enable conclusions on what tools the online brand owners have to ensure positive online brand experience for their consumers and what online brand experience building blocks are necessary to succeed in this process.

Analyzing the online brand experience creation models several studies are worth to be scrutinized in details. Wood and Masterman proposed the “7I” model, which enhances the event experience [47]. A. T. Kearney developed the “7C” online brand experience creation model, which is based on the McCarthy “4P” marketing mix tools [29, 33]. Constantinides suggested the model of three online brand experience building blocks, focused mainly on the web applications [13]. Smith and Wheeler propose people, product and services as the main tools which lead to brand experience [43].

Wood and Masterman proposed “7I” event experience enhancement model although originally focusing on the event marketing is also relevant to this study [47]. Web site visits can be analyzed as the event case. Proposed model shall broaden the overall understanding of the online brand experience and event experience can be assumed as consumer or even consumption experience [46]. For it the main factors are: (i) the brand’s ability to involve consumers emotionally; (ii) brand’s interaction with consumer; (iii) all senses of immersion (deepening of sensual experiences); (iv) memorable intensity; (v) customized individuality; (vi) creative innovation, which encompasses web site content; (vii) sound and authentic integrity providing the value to the consumer [46]. The high consumer involvement can be achieved by the attractiveness of the brand, additional benefits besides product or services consumption or the value which consumer sees as a “cherry on the cake”. This “cherry” is the integrity factor, ensured through the additional value perceived by the consumers. Interaction provides conditions for the networking with other people, who are brought together by the brand and jointly share individual experience, gained through the customized approach to the consumer engagement by the brand.

A. T. Kearney “7C” model designated (i) content, (ii) convenience, (iii) customer care, (iv) communication, (v) connectivity, (vi) community, and (vii) customization as the main tools “<…to deliver a tangible…experience in the virtual space>” [29, p. 10]. The author of this model highlights that content must be relevant and useful for the consumer. It shall be closely related to product or service offering (accurate product description) in order to compete with the offline business. Describing the convenience Kearney is noticing that easy navigation and speed of the web site are essential, for the consumers not to waste their time waiting for it to load [29].



Communication has to ensure intensive interaction with the consumer, two way dialogue in a way which is tailored for exact consumer. Consumer care is inconceivable without guaranties and assistance in every step. Also it is closely related to security risks and issues, such as the violation of the consumer’s data or privacy and result in the consumers mistrust. Connectivity is created through the affiliate marketing tools. They enable the brand to meet its target audience efficiently throughout the collaborators web sites. Building community brings people together to express their opinions, perceptions and experience about particular brand. This factor fulfillment enhances brand’s boundaries and when carefully monitored it builds stronger brand experience. A. T. Kearney also suggests customization of the online experience for the consumer by creating special product offering [29].

Constantinides also provides the rich study of factors influencing online experience. He constructs three main building blocks from (i) functionality, (ii) psychological factors, and (iii) content (iii) [13]. Each of these blocks is divided into bricks, and implements whole online brand experience creation process. Web site functionality is divided into usability and interactivity. Usability creates user-friendly web site, while interactivity presents consumer with grater web experience by communicating as brand-consumer, and consumer-consumer (chance to build a community). Psychological element encompasses online trust creating tools, which for pure-plays are the main focus for designing the web site [13]. Content block proposed by Constantinides consists of the aesthetic and marketing elements presented in the web site. Aesthetics in this model is seen as critical driver for consumer purchasing decision making process. Traditional “4P” marketing mix with its elements is augmented by Constantinides with such additional drivers as communications (to cover lack of physical contact), fulfillment of the orders and product or service characteristics.

Smith and Wheeler associate (i) people, (ii) product and services, (iii) process to the online brand experience building process [43]. In order to explain this model Smith provides triangle of main factors influencing online brand experience (see Figure 1):

FIGURE I: THE BRAND MANAGEMENT ICEBERG. SOURCE: SMITH, WHEELER, 2002, P. 99

People

Proposition

Products

Process

People

Managing the experience

Managing the expectation



www.elvedit.com 108

Smith suggests to start from the clear page, but at the same time differentiating from competitors propositions, which brings precise idea of what and how should be delivered to consumer [42].

This idea will help to manage consumer expectation and set the directions for comprehensive approach to brand experience creation process. The main tool for that, according Smith, is People. It requires a lot of investment to the employees, especially those who are communicating with the consumers directly, but such investments are paying dividends by bringing company’s employees as brilliant brand ambassadors, who create differentiated experience for the consumers of the brand. Electronic marketing tools such as customer relationship management (CRM), voice-activated response systems or direct e-mails do not assist in creating experience, if they are poorly managed by the business personnel (people) [40]. When describing elements of the process, Smith [42] brings as an example Amazon “One-Click” ordering solutions, which highly enhanced consumers satisfaction by shortening the time spent on the ordering process. These kinds of tools implement given brand promises and create memorable experience, which ensures consumers trust and loyalty.

Products or provided services influence overall brand experience, but the degree of influence significantly depends on the differentiation and the value that the product or service of the business provides. If extra benefit is provided for the consumer, he or she is going to be highly involved in brands existence. That requires contemplation on the deliberate ways on how to create unique product/ service experience, which shall include not only products characteristics or packing design, but also such elements as online ordering systems, automated e-mails, return policies and other tools.

Main analyzed online brand experience creation models are summarized below for the comparative analysis (see Table 2).

All compared models can be seen as a continuation of the traditional marketing mix implemented in online environment with an aim to create comprehensive online brand experience. Traditional “4P” + “1P” (People) model [42] is transformed into “7I”, “7C”, “3P” or “3 building blocks”. All of them include requirements for products/ services and place – web site. Price is excluded from almost all of the models as not essential element in online brand experience [13]. According Constantinides [13] promotion also can be excluded from the list of essential online brand experience factors. All of these models can be employed in creating the online brand experience and they provide some guidance for the businesses trying to reach better results in online environment, but they fail to emphasize the digitization of the brand itself.



TABLE II: ONLINE BRAND EXPERIENCE CREATION MODELS. SOURCE: COMPILED BY THE AUTHORS.

Wood & Masterman (2007) “7I”

model

A. T. Kearney (1999) “7C” model

Constantinides (2004) online brand experience “3 building blocks”

Smith and Wheeler (2002) “3P” model

Individuality

Convenience Ease of use Navigation Speed of delivery Dependability

Functionality factors

Process How can our processes deliver our products and services in a valuable way?

Usability: Interactivity: - Convenience - Site navigation - Information architecture - Ordering/payment processes - Search facilities and process - Site Speed - Find-ability/ accessibility

- Customer service/after sales - Interaction with company personnel - Customization - Network effects

Interaction

Customization Ability to tailor interface Product offering

Communication Integrated two way dialogue Visual presentation and aesthetics.

People What must be distinctive about our people

Integrity Real benefits and value to the consumer

Customer care Integrated customer management: - Contact management - Pre-/post-sales support service strategy

Psychological factors (trust) Transaction security Customer data misuse Customer data security Uncertainty reducing elements Guarantees/return policies

Involvement

Immersion Of all senses Community

Innovation Content Content factors

Aesthetics: - Design - Presentation quality - Design elements - Style/atmosphere

Marketing mix: - Communication - Product - Fulfillment - Price - Promotion - Characteristics

Product/Service Offering What must be unique about our products and services?

Intensity

Connectivity Site to site connectivity: - High quality related links - Leveraging of alliance sites

IV. DESIGNING ONLINE BRAND EXPERIENCE CREATION PROCESS MODEL The comparative analysis of online brand experience creation models enables us to depict most

important factors in the process of building the effective experience for the brand consumer in the online environment. All analyzed models can be combined in order to reach integrated holistic



www.elvedit.com 110

improvement in online brand experience building process accounting for the digitization of the brand itself (see Figure 2).

FIGURE II: ONLINE BRAND EXPERIENCE CREATION PROCESS MODEL. SOURCE: COMPILED BY THE AUTHORS.

Proposed model covers most predominant online brand experience building blocks – product/services, place, which is brand/company’s web site, people, who are brand ambassadors, and online process, which is essential for all building blocks of the brand experience. Product and place alone ensures basic consumer needs, but only in combination with the well managed online process they can provide sophisticated online brand experience.

Product/service element, which enhances online brand experience, means proper presentation, providing of the assortment, which shall meet to the advertised characteristics and consumer needs, but at the same time shall assume that it is customizable for a particular individual. Products and service of the brand must integrate into consumer life [48] and help to express consumer itself. The more it is self-expressive, the more it can involve consumer to be converted into the brand advocate [45]. Marketing practitioners also suggest exploiting multi-sensorial elements while presenting product/services offerings [11]. It is quite difficult to implement all five senses in online environment, but at least sound and visual elements shall be exploited accurately in order to support brand’s message. Also the other senses shall be stimulated indirectly using appropriate visual elements, which can evoke consumer’s memories about the smells and tastes (for example: picture of a lemon or the smell of cut grass). Kotler also mentions the virtual reality capabilities “<…that allows users to experience three-dimensional, computer

PRODUCT/ SERVICE

Characteristic Individuality Information Immersion

ONLINE PLACE Content

Customization Convenience

Site navigation Site speed

Connectivity Innovation

PEOPLE Consumer care Communication

Community Service

Interaction

ONLINE PROCESS Integrity / Intensity / Involvement

Transaction security Consumers data security

Speed of delivery/ fulfillment of orders Uncertainty reducing elements/guarantees

ONLINE BRAND EXPERIENCE



generated environments through sound, sight, and touch>” [31, p. 150]. These tools help enhance the consistency of proposition and create holistic online experience, which lets the consumers know what the brand stands for and what benefits or values will be delivered to them every time that they are interacting with it.

Place is where consumers meet the brand. It can be either web site of the business or the brand, or other online place of affiliated partners, communities, social networks, where consumer can be engaged by the brand or at least can get know the brand. All of these places need to ensure well managed brand and consumer interaction points. Content, customization, convenience, site navigation and speed, connectivity (search facilities, accessibility) and innovation are most predominant factors in all proposed models presented before. According to Treiblmaier research content is a “<…major success factor and has a great influence on how customers perceive a company and rate the e-brand>” [44, p. 92]. Customization differs from individuality (Product element) by letting the consumer to decide and pick his/her own preferences, while individuality is automatically set by the web site content offering. Convenience, site navigation and speed are about letting the consumer to use web site easily. Search facilities and accessibility describe web site connectivity. They enable the consumers to find online brand easily in the environment overcrowded with information. It is always a challenge to choose between web site design, aesthetics and usability while ensuring all the factors listed above. But it is necessary to ascertain, what the users of that web site (place) demand from particular brand and to remember that experience can be gained only if the consumer can make use of the offerings of that web site. Thus, usability should be considered as paramount to the implementation of creative and aesthetic solutions. When deciding what usability tools have to be implemented, intuitiveness factor should be taken into account. Consumers are not willing to spend their time on long learning on how to use the web site. Thus, the technology used to support the web site has to include the functions of auto-complete, on-site search, with high degree of typing errors tolerance [37]. All elements and tools described above must be revised constantly and innovated to create the better online brand experience for their consumers.

Place as one of the building blocks of online brand experience is only part of the rationalized online brand experience. To provide emotional benefit to consumer and to create greater experience, a third block – people, has to be used. People can be divided into two groups – people, who are employed to be brand ambassadors, and another group – people, who love the brand and are connecting in online communities or in other places provided by the social media channels (brand evangelists). The first group of people delivers consumer care, communicates with them to ensure well managed relationship, keeps interacting with consumers by two way dialogue, mentors brand experience within the communities and provides all necessary service to enhance online brand experience [6].

Service is one of the main factors influencing consumer’s satisfaction [10]. This element demands fast reaction to consumer needs in order to ensure the quality of experience [32]. Communication in Treiblmaier study is presented also as one of the most important elements. Treiblmaier distinguishes three forms of possible communications: (a) cross-media



www.elvedit.com 112

communication, (b) pure online communication, and (c) pure offline communication [44]. Cross-media communications are based on combination of online and offline media capabilities. Pure online communication is more used by pure online players [44], however pure offline communication is no longer suitable for both categories of companies – online and also for traditional bricks and mortar businesses, especially if they have intentions to shift part of their business to online environment. Communication element can be compensated by passive information to a certain degree, but it cannot totally reduce consumer’s uncertainties [13]. This can be done by providing guaranties, taking care of consumer data and transaction security, providing excellent fulfillment of the orders. Second group of people, in the people building block, are the consumers themselves. The people engaged by brand and gathered together are brand evangelists next to the people employed by the brand owners. Their provided reviews are “<…16 times more trusted than other reviews>” [27, p. 9]. Consumers interact with each other and share their perceptions and by this way enhance the overall online experience of the brand. Also consumers need to socialize has to be considered as an important factor. It draws attention to social media and its marketing tools offered by it in order to engage consumers. If a brand provides this opportunity by setting at least the fan page in the Facebook or Twitter, in this way it ensures that brand interaction with consumer is more intense and brings consumers into social conversation. Social network interaction is also an option for seeking consumer’s feedback on the brand, ensuring deliberate communications and reinforcing the consistency of brand experience. Everything what is happening in the social communities or social media fan pages needs be monitored by the business in order to actively manage and respond to it. This way the consumer interaction returns to the first group of people – brand ambassadors, as the building block of the online brand experience, which manage the correct selection of marketing tools and implement the strategy for creation of positive online brand experience.

The fourth online brand experience creation block is the online process element. It consist of such sub-elements as integrity, intensity, involvement, transaction security, consumer’s data security, speed of delivery, fulfillment of orders, uncertainty reducing elements and guaranties. Note that the challenges and processes online are very different that in offline bricks and mortar business environment. Some of these sub-elements were already explained as the online brand experience building blocks above. The online process element is all-inclusive element which filters the products/services, online place and people elements (“what” elements) implementation turning into the sophisticated effective and holistic experience. It answers the questions “how”, “when” and “what” should be provided to the consumer in order to create holistic online experience. It ensures that various marketing tools are implemented completely and consistently. This element if effectively employed can become the primary differentiator means for online brand owners. Holistic online brand experience can be ensured by making the web site relevant for the consumers, offering integrity, intensive involvement and reliability. Additionally holistic online brand experience means proactive management of the consumer engagement at every interaction point. Already existing analytics engines provides the ability to monitor how successfully online process is implemented and decide what should be improved in order to enhance positive online brand experience for the consumers.



Proposed online brand creation process model is conceived by combining different existing models and focusing them on the online event brand experience management and web-site brand experience management. Although the main building blocks, fulfilled by appropriate sub-elements, are identified, further research is necessary to validate and reshape them for the dynamic online environment. Based on the central role of the online process the proposed model may be nicknamed - 3PoP (Production, Place, People, online Process).

Proposed model can be useful not only for the pure online businesses, but also for offline businesses. Although online brand experience can be created by the similar building blocks as the brand creation in the traditional environment, however the touch points of the brand with the consumers are distinct and specific in the online environment, therefore their implementation process is going to be different. The 3PoP model proposed in this paper helps account for these differences at both research and management levels. It is necessary to mention that such complexity requires contribution of all internal business processes [18]. It cannot be implemented only through marketing measures [42]. The comprehensive 3PoP marketing approach shall ensure the consistency of overall holistic brand experience online.

V. CONCLUSIONS AND RECOMMENDATIONS 1. Positive brand experience provides conditions for brand familiarity, satisfaction and

influence further purchases. Consumer’s loyalty and trust are strong positive outcomes of the online brand experience. In order to reach these outcomes in e-business, multiple digital interaction channels need to be engaged, merged and managed. Memorable, intensive and integrated experience for the consumers creates differentiation of online brand and ensures top positions among similar brands. This requires thorough understanding and analysis framework of the online brand experience concept and the process of creation thereof. Review of existing literature provides extensive brand experience conceptualization, however it is not sufficiently applied to the online environment. The dominant characteristics of the brand experience include internal and external consumer responses, evoked by brand-related stimuli. Internal responses include sensations, feelings, and cognitions, while external responses assert in consumer behavior. This is a commonality of both offline and online brand experience concepts. Thus, the online brand experience definition proposed in the paper is focused on consumer perceptions and responses to brand evoked stimuli.

2. Our analysis of the process of the online brand experience creation concludes that most existing online brand experience creation models are restated continuation of traditional “4P” marketing complex. In the existing research products/services and place are main factors influencing online brand experience. Other essential brand experience building blocks are the people and the marketing process. Analysis suggests that specific online processes and their role are insufficiently reflected in the existing models. We argue that the online processes are central for the holistic online marketing mix and online consumer experience. Based on this, the paper proposes a modified online brand experience creation model - 3PoP, which is delineated by the main building blocks:



www.elvedit.com 114

• Product/service – the object (or “what”) which provides to consumers the fulfillment of rational and emotional needs;

• Place improved by the online marketing tools is where strong online brand experience is felt;

• People as the brand mediators delivering online brand experience. This includes two groups of people – people, who are employed by the brand owners (brand ambassadors), and people who are in love with the brand and gather together into online communities to share their experience and to enhance other brand consumer experience by provided reviews and other responses (brand evangelists);

• Online process - the main filter and transforming factor for the product/service, place and people into the holistic online brand experience. It defines “how” and “when” something should be done to ensure completeness and consistency of online brand experience.

3. The proposed model is based on theoretical insights from the meta-analysis of the current scientific literature and attempts to merge the main factors suggested in preceding research. Proposed modified online marketing model focused on the online brand experience creation shall be useful for online-businesses, as well as for offline businesses. Identified main building blocks are essential for better understanding and strategic management of online brand experience creation. Implementation of these blocks needs to be customized for the specific business environment and needs to be considered as a comprehensive complex. The contribution of all business processes (not only the marketing process) is necessary in order to ensure holistic online brand experience. Further research, empirical studies and case studies are needed in order to validate the proposed 3PoP model, along with the continuation of the theoretical discussion on the creation of the online brand experience.

REFERENCES [1] Aaker J. L. (1997). Dimension of Brand Personality. Journal of Marketing Research, Vol. 34 (August),

347–56.

[2] Atwal G., Williams A. (2009). Luxury brand marketing — The experience is everything! Journal of Brand Management, Vol. 16 No. 5/6, 338 – 346.

[3] Barwise P. (2003). Preface. In Clifton et all. 2003. Brands and Branding. The Economist. Retrieved from<http://appsler.com/186197664X.pdf#page=114>.

[4] Blackett T. (2003). What is a brand? In Clifton et all. 2003. Brands and Branding. The Economist. Retrieved from: <http://appsler.com/186197664X.pdf#page=114>.

[5] Brakus J. J., Schmitt B. H., and Zarantonello L. (2009). Brand Experience: What Is It? How Is It Measured? Does It Affect Loyalty? Journal of Marketing, Vol. 73, 52 – 68.

[6] Brodie R. J., Whittome J. R. M., Brush G. J. (2009). Investigating the service brand: A customer value perspective. Journal of Business Research, Vol. 62, 345 – 355.

[7] Brymer Ch. (2003). What makes brands great. In Clifton et all. 2003. Brands and Branding. The Economist. Retrieved from: <http://appsler.com/186197664X.pdf#page=114>.



[8] Carpenter P. (2000). EBrands: Building an Internet Business at Breakneck Speed. Harvard: Harvard Business School Press Books.

[9] Chaffey D., Mayer R., Johnston K., Ellis-Chadwick F. (2000). Internet Marketing. Harlow: FT Prentice Hall.

[10] Chaffey D., Ellis-Chadwick F., Mayer R., Johnston K. (2006). Internet Marketing: Strategy, Implementation and Practice, 3rd edition. Pearson Education Limited.

[11] Charlton Ch. (2013). Find success through great online brand experiences. Retrieved from: <http://www.equestriancreativenetwork.com/news-stories/find-success-through-great-online-brand-experiences-1206>.

[12] Christodoulides G., de Chernatony L. (2004). Dimensionalising on- and offline brands composite equity. Journal of Product & Brand Management, Vol. 13, No. 3, 168 – 179.

[13] Constantinides E . (2004). Influencing the online consumer’s behavior: The Web experience. Internet Research, Vol. 14, No. 2, 111 – 126.

[14] Davidaviciene V., Gatautis R., Paliulis N., Petrauskas R. (2009). Electronic business. Vilnius: VGTU.

[15] De Chernatony L. (2000). Succeeding with brands on the internet. Birmingham Business School Working Paper Series, p. 186 – 195. Retrieved from: <http://www.brandchannel.com/images/papers/success.pdf>.

[16] De Chernatony L. (2001). A model for strategically building brands. Brand management, Vol. 9, No. 1, 32 – 44.

[17] De Chernatony L., Dall’Olmo Riley F. (1999). Experts’ Views about Defining Services Brands and the Principles of Services Branding. Journal of Business Research, Vol. 46, No. 2, 181 – 192.

[18] De Chernatony L., Horn S. (2003). The criteria for successful services brands. European Journal of Marketing, Vol. 37, No. 7/8: 1095 – 1118.

[19] Endelman D. C. (2010). Branding in the digital age: You’re Spending Your Money in All the Wrong Places.Harvard business review, December. Retrieved from: <http://www.google.lt/url?sa=t&rct=j&q=edelman%2C%20d.c.%20branding%20in%20the%20digital%20age&source=web&cd=2&sqi=2&ved=0CCcQFjAB&url=http%3A%2F%2Fwww.edmarbulla.com.br%2Fdownload%2Fartigos_de_marketing%2FBranding%2520in%2520the%2520Digital%2520Age.pdf&ei=6FxLUIz4NIzPsgaUx4HICQ&usg=AFQjCNEfw5UH2h1tl9URbFbugDjqG5Qn4w>.

[20] Ha H. Y, Perks H. (2005). Effects of consumer perceptions of brand experience on the web: Brand familiarity, satisfaction and brand trust. Journal of Consumer Behavior, Vol. 4, No. 6, 438 – 452.

[21] Hoffman D. L., Novak T., Peralta M. (1999). Building consumer trust in online environments: the case for information privacy. Communications of the ACM, Vol. 42, No. 4, 80 – 85.

[22] Ibeh K. I. N., Luo Y., Dinnie K. (2005). E-branding strategies of internet companies: some preliminary insights from the UK. Journal of Brand Management, Vol. 12, No. 5, 355 – 373.

[23] Iglesias O., Singh J. J. (2011). The role of brand experience and affective commitment in determining brand loyalty. Journal of Brand Management, Vol. 18, 570 – 582.

[24] Jack Morton Worldwide global brand experience agency(2011). “Best Experience Brands” a global



www.elvedit.com 116

study. Retrieved from: <http://www.krcresearch.com/pdfs/BestExperienceBrands.pdf>.

[25] Jevons C., Gabbott M., de Chernatony L. (2000). Internet noise and its effects on brand management. December 22. Retrieved from: <http://www.brandchannel.com/images/papers/noise_final.pdf>.

[26] Kahn B. (2011). A Framework for Understanding the Relationship between Advertising and Retailing. In Future of Brand Building and Brand Experience: The Blurring Boundaries between Advertising and Retailing conference report prepared by Finnel S. Retrieved from: <http://www.wharton.upenn.edu/bakerretail/files/baker-future-of-brand-building-conference-report.pdf>.

[27] Kantor B. (2011). Making the Most of Social Networking and Online Media. In Future of Brand Building and Brand Experience: The Blurring Boundaries between Advertising and Retailing conference report prepared by Finnel S. Retrieved from: <http://www.wharton.upenn.edu/bakerretail/files/baker-future-of-brand-building-conference-report.pdf>.

[28] Kapferer J. N. (2012). The new strategic brand management: advanced insights and strategic thinking, 5-th ed. London: Kogan Page.

[29] Kearney A. T. Global management consultancy and executive search firm(1999). Creating a High-Impact Digital Customer Experience. White paper. Retrieved from: <http://www.labeee.ufsc.br/~luis/egcec/artigos/atk-digital%20customer.pdf>.

[30] Klaus P., Maklan S. (2007). The role of brands in the service-dominated world. Brand management, Vol. 15, No. 2,115-122.

[31] Kotler P. (2002). Marketing Management, Millennium Edition. Pearson Custom Publishing. USA, Boston.

[32] Lee G. G., Lin H. F. (2005). Customer perceptions of e-service quality in online shopping. International Journal of Retail & Distribution Management, Vol. 33, No. 2, 161 – 176.

[33] McCarthy E. J. (1964).Basic marketing a managerial Approach. Irvin, Homewood, IL.

[34] Morrison M., Crane F. G. (2007). Building the service brand by creating and managing an emotional brand experience. Journal of Brand Management, Vol. 14 No. 5, 410 – 421.

[35] Najmaei A., Sadeghinejad Z. (2009). E-Buyology - a Step towards I-Branding: a Holistic Conceptual Paradigm. Creating Global Economies through Innovation and Knowledge Management: Theory & Practice, p. 834–843.

[36] Norton D. W. (2003). Toward meaningful brand experiences. Design Management Journal, Vol. 14, No. 1, 19 – 25.

[37] Pusch A. (2011). How to improve your brand experience online. Retrieved from: <http://www.utalkmarketing.com/Pages/Article.aspx?ArticleID=21024&Title=How_to_improve_your_brand_experience_online>.

[38] Ryan D., Jones C. (2009). Understanding Digital Marketing. Marketing Industry Network, Kogan Page, Glasgow.

[39] Sahin A., Zehirb C., Kitapçı H. (2011). The Effects of Brand Experiences, Trust and Satisfaction on Building Brand Loyalty. Procedia Social and Behavioral Sciences, Vol. 24, 1288 – 1301.



[40] Simmons J. (2007). “I-Branding”: developing the internet as a branding tool. Marketing Intelligence & Planning, Vol. 25, No. 6, 544 – 562.

[41] Sinha I. (2000). Cost Transparency: The Net’s Real Threat to Prices and Brands. Harvard Business Review. March–April. Retrieved from: <http://faculty.msb.edu/homak/homahelpsite/webhelp/Cost_Transparency-_Nets_Real_Threat.pdf>.

[42] Smith S. (2003). Brand experience. In Clifton et all. 2003. Brands and Branding. The Economist. Retrieved from: <http://appsler.com/186197664X.pdf#page=114>.

[43] Smith S., Wheeler J. (2002). Managing the customer Experience: Turning customers into advocates. Executive summary. Managing the Customer Experience. FT Prentice Hall. Retrieved from: <http://www.smithcoconsultancy.com/wp-content/uploads/2008/07/mce-execsummary.pdf>

[44] Treiblmaier H. (2006). Determinants of Electronic Branding: An Exploratory Study. Innovative Marketing, Vol. 2: 89 – 98.

[45] Wallace E., Buil I., de Chernatony L. (2012). Facebook ‘friendship’ and brand advocacy. Journal of Brand Management, Vol. 20, No 2: 128 – 146.

[46] Wood E. H. (2009). Evaluating Event Marketing: Experience or Outcome? Journal of Promotion Management, Vol. 15: 247 – 268.

[47] Wood E. H., Masterman G. (2007). Event Marketing: Measuring an experience? 7th International Marketing Trends Congress – Jan 17th/19th Venice. Retrieved from: <http://www.escp-eap.net/conferences/marketing/2008_cp/Materiali/Paper/Fr/Wood_Masterman.pdf>.

[48] Wright P., McCarthy J., Meekison L. (2004). Making sense of Experience. In Blythe et all. Funology, p. 43-53. Retrieved from: <http://www.cos.ufrj.br/~jano/CSCW2008/Papers/Funology%20Introduction.pdf>.

[49] Zarantonello L., Schmitt B. H. (2010). Using the brand experience scale to profile consumers and predict consumer behaviour. Journal of Brand Management, Vol. 17, No. 7, 532 – 540.


Tadas Limba is a Head of the Institute of Digital Technologies at Mykolas Romeris University in Vilnius, Lithuania. He got B. Sc. in Politics from Vilnius University, 1999 and B. Sc. in Law from Mykolas Romeris University, 2010. He got M. Sc. in Public Administration from Mykolas Romeris University, 2001 and M. Sc. in Business Law from Mykolas Romeris University, 2012. Tadas Limba also got his Ph. D. in Management and Administration from Mykolas Romeris University, 2009. Tadas Limba is an Associate Professor from 2010.

Tadas Limba has published over 20 articles in Lithuanian and foreign scientific journals, monograph, textbook, focused on e-government and e-business. His additional areas of research and expertise are – IT law regulation and policy; digital content and digital media, privacy and data protection issues. Tadas Limba is a member of Lithuanian Computer Society since 2007. Since 2013 he is Asia Center Board Member, South Korea's representative at Mykolas Romeris University. He is visiting professor at Zaragoza University in Spain. He plays an active role in international communication and development of joint double degree studies program with South Korea Dongseo University. Tadas Limba made presentations in



www.elvedit.com 118

international and national conferences. Tadas Limba is fluent in English, Spanish and Russian, he is also elementary user of German.

Prof. Mindaugas Kiškis is the Professor at the Mykolas Romeris University in Vilnius, Lithuania. Prof. Kiškis is main areas of research and expertise are – intellectual property, technology (bio-nano-ICT) management, regulation and policy; digital content and digital media, privacy and data protection issues. Prof. Kiškis also works in e-business, innovation and entrepreneurship fields. Mindaugas Kiškis holds the PhD (2002) from the Mykolas Romeris University, graduate

degrees from the Vilnius University (1998) and the Baltic Management Institute (Vytautas Magnus University) (2005). Prof. Kiškis is a fellowships recipient with major foreign universities, including Visiting Professor at the East China University of Science and Technology (2013), Understanding Canada Scholar (2011), Fulbright Scholar at the Arizona State University (2007-2008), Markle Fellow at the Oxford University (2003) and other. Prof. Kiškis is also the serial entrepreneur, cofounder of three technology startups, active member of the knowledge society NGOs, also, the author of 4 monographs, 4 textbooks and 35 articles in Lithuanian and foreign publications. Dr. Kiškis is fluent in English and Russian, he is also elementary user of Mandarin. More info at prof. Kiškis blog: www.kiskis.eu

Virginija Jurkutė got B. Sc. in Public Administration at the General Jonas Žemaitis Lithuanian Military Academy in 2007. She got M. Sc. in Electronic Business Management at Institute of Digital Technologies, Faculty of Social Technologies, Mykolas Romeris University, Lithuania in 2013. Her research interests: electronic marketing, branding in electronic environment, electronic commerce, electronic business, public relations. Virginija Jurkutė also serves in Lithuanian armed forces.


Color Image Segmentation Using a Modified Fuzzy C-means Method and Data Fusion

Techniques

Authors

Rafika Harrabi ESSTT/CEREP Unit/ University of Tunisia

[email protected] 5 Av. Taha Hussein,1008, Tunis

Ezzedine Ben Braiek ESSTT/CEREP Unit/ University of Tunisia

[email protected] 5 Av.Taha Hussein,1008, Tunis

Abstract

In this paper, a new color image segmentation method based on modified Fuzzy c-means and data fusion techniques is presented. The proposed segmentation consists in combining many realizations of the same image, to gether, in order to increase the information quality and to get an optimal segmented image. In the first step, the membership degree of each pixel is determined by applying fuzzy c-means clustering to the information coming from the component images to be combined. The idea is to link at the image pixel level, the notion of measurement functions to that of membership functions in fuzzy logic. In the second step, the fuzzy combination theory is employed to merge the component images of the original image, in order to increase the quality of the information and to obtain an optimal segmented image. Segmentation results from the proposed method are validated and classification accuracy for the test date available is evaluated, and then a comparative study versus existing techniques is presented. Experimental segmentation results of color medical and textured images show the effectiveness of the proposed method.

Key Words

Segmentation, biomedical image, fuzzy c-means, fuzzy fusion, conflict, data fusion.

I. INTRODUCTIONThe image segmentation is an essential component which determines the quality of the final

results and analysis [1] [2]. It consists in partition of an image into homogeneous regions, according to a choice criterion, such as intensity, color, tone or texture, etc.

Recently, color image segmentation attracts more and more attention [3]. The situation often



www.elvedit.com 120

occurs when the objects cannot be extracted using gray scale information but can be extracted using color information. Compared to gray scale, color provides additional information to the intensity. Most gray level image segmentation techniques such as histogram thresholding, clustering, region growing, edge detection, fuzzy methods, and neural networks can be extended to color images. Gray level segmentation methods can be applied directly to each component of a color space, and then the results can be combined in some way to obtain a final segmentation result [4].

Many different techniques of color image segmentation have been developed and detailed in the literature [5]. The general segmentation problem consists in choosing the adopted color model for a specific application.

In the Red, Green, Blue (RGB) representation, the color of each pixel is usually represented on the basis of the three primary colors (red, green, and blue), but it can be coded in other representation systems. Different representation systems have been developed by several authors [6] [7]. These spaces are obtained by using the linear and non-linear transformations of the RGB color space. Each color representation has its advantages and disadvantages. There is still no color representation that can dominate the others for all kinds of color images yet.

In the past, many authors have addressed the color image segmentation problems using different methods [2] [3] [4], and several researchers have, in particular, investigated the data fusion techniques and fuzzy logic [8] [9]. Preliminary works using fuzzy techniques such as Fuzzy C-Means (FCM) [10] and Hard C-Means (HCM) algorithms [11] have also been reported in literature.

However, FCM algorithm [10] has a considerable difficulty in noisy environments, and the memberships resulting from this algorithm do not always correspond to the intuitive concept of degree of belonging or compatibility. The membership degrees are computed using only gray levels and do not take into account the spatial information of pixels with respect to one another. Also, the Hard C-Means (HCM) [11] is one of the oldest clustering methods in which HCM memberships are hard (i.e., 1 or 0).

In this context, Liew et al. [12] have provided a new dissimilarity index that considers the influence of the neighboring pixels on the centre pixel in the FCM algorithm. With the same objective, Ahmeh et al. [13] have introduced a regularization term in the standard FCM algorithm to impose the neighborhood effect. However, typical problems remain difficult to solve including the more consuming time is needed during the computation.

Recently, data fusion techniques have been tested for medical image segmentation [14] [15]. Data fusion is a technique which simultaneously takes into account heterogeneous data coming from different sources, in order to obtain an optimal set of objects for investigation. In the past, many authors have addressed this problem using different methods [4] [14] [15].

In this context, Vannoorenberghe et al. [14] have proposed an information model obtained from

Color Image Segmentation Using a Modified Fuzzy C-means Method and Data Fusion Techniques Rafika Harrabi and Ezzedine Ben Braiek


training sets extracted from the pixel intensity of the image. In their papers, the authors described the estimation of the Model Mass Function method based on the Assumption of Gaussian Distribution (MMFAGD) and histogram thresholding and applied on synthetic and biomedical images that contain only two classes.

In another study, Zhu et al. [15] have proposed a segmentation method based on fuzzy sets and Dempster-Shafer (DS) evidence theory. The idea is to assign, at each image pixel level, a mass function that corresponds to a membership function in fuzzy logic. The membership degree of each pixel is determined by applying the FCM algorithm to the gray levels of the image. Then, the DS combination rule and decision are applied to obtain the final segmentation.

With the same objective, Ben chaabane et al. [4], have proposed a color image segmentation method based on fuzzy homogeneity and Dempster-Shafer evidence theory. The fuzzy homogeneity vector is used to determine the fuzzy region in each primitive color, whereas, the evidence theory is employed to merge different data sources in order to increase the quality of the information and to obtain an optimal segmented image.

In this paper, we present a new color image segmentation method based on a modified fuzzy c-means algorithm and data fusion techniques, applied to a specific kind of medical image segmentation, where we aim at providing a help to the doctor for the follow-up of the diseases of the breast cancer. The problem of color image segmentation is addressed using the fuzzy fusion theory. The objective is to rebuild each cell from a series of images. These images are fused together by the fuzzy fusion theory using as input features, the membership degree of each information extracted from these input images, previously estimated and associated to each pixel. The modified fuzzy c-means is used to calculate the membership degree of each pixel. Once the membership degrees are estimated for each image to be fused, the fuzzy combination rule and decision are applied to obtain the final segmentation. Consequently, the proposed algorithm uses a centralized fusion model that requires the availability of all the images simultaneously, and no intermediate decision is taken before fusion.

Section 2 introduces the proposed method for color image segmentation. The experimental results are discussed in Section 3, and the conclusion is given in Section 4.

II. THE PROPOSED METHOD Image segmentation consists in partition of an image into homogeneous regions. In the

framework of our application, we are interested in color image segmentation of cells in the breast images. In fact, the problem is to separate the cells from the background. To do this, the fuzzy c-means (FCM) algorithm can be used to determine the membership degree of each pixel, but it is based on only gray level and does not take into account the spatial information of pixels with respect to each other. Consequently, the statistical features extracted from the component images (H, S and I) to be combined can be used to overcome this drawback. The fuzzy c-means algorithm technique is used to extract homogeneous regions, in each component images of the original image expressed in the HIS color space. Once the membership degrees are estimated, the fuzzy



www.elvedit.com 122

combination rule is applied to obtain the final segmented image. Hence, the main idea of the proposed method is to fuse, one by one, the pixels of the input images.

However, to fuse different images using fuzzy fusion theory, the appropriate determination of membership degrees plays a crucial role, since assignation of a pixel to a cluster is given directly by the estimated membership functions. In the present study, the method of generating the membership functions is based on the modified fuzzy c-means algorithm. To do this, the first order statistical feature vector is included in the Fuzzy C-means technique to provide memberships degrees witch correspond to the intuitive concept of degree of belonging or compatibility.

A. Determination of membership degrees

The standard Fuzzy C-means algorithm (FCM) [10], an unsupervised clustering algorithm, has been applied successfully to a number of clustering problems. The algorithm minimizes the objective function for the partition of data set, s

d RxxxX ,...,, 21 , given by:

),(),( 2

1 1ik

d

k

c

i

mikm vxduvuJ

(1)

with

djuc

iij

1

1,1 (1a)

djciuij 1,1,0 (1b)

n

jij ciu

1

1,0 (1c)

where sd RxxxX ,...,, 21 , s is the dimension of space, 푑 is the number of samples, c is the

number of clusters dc 1 , m is the fuzzy factor 1m , ijij vxd is the distance between the

sample jx and clustering center iv , si Rv with ci 1 . iju is the membership of the jth sample to

the ith clustering center, ijuU is a matrix of size (cd). cvvvV ,...2,1 is a matrix of size (cs).

The FCM algorithm minimizes the objective function ),( vuJ m with respect to the membership functions jku and the centroids kv . The FCM clustering technique can be summarized by the following steps:



However, this algorithm has a considerable drawback in noisy environments, and the memberships degree resulting from FCM do not correspond to the intuitive concept of belonging or compatibility. Instead, the FCM relaxes the condition, and allows the feature vector to have multiple membership grades to multiple clusters, Suppose the data set with known clusters and a data point which is close to both clusters but also equidistant to them. Fuzzy clustering gracefully copes with such dilemmas by assigning this data point equal but partial memberships to both clusters, that is the point may belong to both clusters with some degree of membership grades varies from 0 to 1.

An image can be represented in terms of pixels, which are associated with a location and a gray level value. It can also be represented by its derivatives, e.g., regions with features like average grayscale value, standard deviation, variance, entropy, third order moment, gradient, etc. These features can be extracted from regions masked by a window (푡 × 푡). The combination of statistical features and standard fuzzy c-means has some advantages for the determination of membership degrees, which correspond to the intuitive concept of belonging or compatibility.

The idea is to replace the vector X used in each image to be combined by a matrix F containing the same number of lines, i.e. (푑 = 푁 × 푀), but with 4 columns. These columns contain 4 statistical features extracted from the sliding window centered around every pixel. Hence, this algorithm scans the image using a (tt) sliding window, as shown in Figure 1, from left to right and top to bottom. A feature vector is extracted from each block.

- Input an (푁 × 푀) image with gray levels zero to 255. Step 1: Initialization (iteration 0) Scan the image line by line to construct the vector X containing all the gray level of the image. Randomly initialize the centers of the classes vectors V(0) From the iteration t=1 to the end of the algorithm: Step 2: Calculate the membership matrix U(t) of element iku using (2a):

c

j

m

jk

ik

ik

d

du

1

1

2

)(

1 (2a)

iku is a matrix of size (cn) Step 3: Calculate the vector V(t) =[v1, v2,…,vc] using:

n

k

mik

n

kk

mik

i

u

xu

v

1

1 (2b)

Step 4: Convergence test: If )1()( tt VV , then increment the iteration t, and return to the step 2, otherwise, stop the

algorithm. is a chosen positive threshold.



www.elvedit.com 124

FIGURE 1: FEATURES EXTRACTION USING A SLIDING WINDOW.

In the present study, the best features are the mean Me, the variance Var, the third order moment Sk and the forth order moment Ku of the window.

Assume g(i, j) is the intensity of a pixel p(i, j) at the location (i, j) in an(N × M) image, w is a size (t × t) window centred at pixel p(i, j).

A feature vector for a pixel is then extracted from the windowed block. The 4 features extracted from the window centered at pixel (i, j) are given by the following equations:

2

1

2

1

2

1

2

1

),(1

t

tk

t

tl

jlikgtt

Me (3)

2

1

2

1

2

1

2

1

2)),((1

t

tk

t

tl

Mejlikgtt

Var (4)

2

1

2

1

2

1

2

1

3)),((1

t

tk

t

tl

Mejlikgtt

Sk (5)

2

1

2

1

2

1

2

1

4)),((1

t

tk

t

tl

Mejlikgtt

Ku (6)



Where (N × M) and g(i, j) are respectively the size of image and the gray scale value of pixel p(i, j), ≤ i ≤ N− and ≤ j ≤ M− . Notes that t must have an odd value to obtain a centered window around each pixel.

So, the modified fuzzy C-means algorithm is used to determine the membership degree of each pixel characterized by 4 statistical features. The proposed modeling information method using the FCM algorithm combined with the statistical features can be summarized by the following steps:

B. Use of fuzzy combination Theory for Image Segmentation

The purpose of segmentation is to partition the image into homogeneous regions. The idea of using fuzzy fusion theory for image segmentation is to fuse one by one the pixels coming from the three images (H, S and I). The modified fuzzy c-means algorithm is applied to the images to be combined. Then, the membership degrees are combined using the fuzzy combination theory to obtain the final segmentation results.

- Input an (푁 × 푀) image with gray levels zero to 255. Step 1: Initialization (iteration 0) Randomly initialize the centers of the classes vectors V(0) of size (푐 × 4) containing the centers of

the classes. Step 2: Compute the matrix F of size (푑 × 4) containing the statistical features extracted from the

image. From the iteration t=1 to the end of the algorithm: Step 3: Calculate the membership matrix U(t) of element iku using (7a):

c

j

m

jk

ik

ik

vF

vFu

1

1

2

)(

1 (7a)

In the modified method, the 퐹 and 푣 are vectors of size (1 × 4). Step 4: Calculate the vector V(t) composed of 4 columns 푣 using:

d

k

mik

d

kk

mik

i

u

Fu

v

1

1 (7b)

Step 5: Convergence test: If )1()( tt VV , then increment the iteration t, and return to the step 2, otherwise, stop the

algorithm. is a chosen positive threshold.



www.elvedit.com 126

TABLE 1. T-NORMS AND T-CONORMS COMBINATION RULES.

T-Norms T-Conorms Name min(x,y) Max(x,y) Zadeh

xy Probabiliste

x if y=1 y if x=1

0 otherwise

x if y=0 y if x=0

1 otherwise

Minimale

max(x+y-1,0) min(x+y,1) Lukasiewicz

Data fusion process consists of combining information from different sources in order to improve the decision process. The goal of data fusion is to reduce uncertainty and imprecision by combining both redundant and complementary data. The different stages of a fusion model are: the definition of measurements 푀, the combination by an operator 퐹 and the decision.

Of the existing data fusion methods such as probability theory [16], fuzzy logic [17] [18], possibility theory [19], evidence theory [4] [20], the fuzzy combination theory [21], is a powerful and flexible mathematical tool for handling uncertain, imprecise, and incomplete information.

In the present study, the information’s comings from different images are represented by the membership degrees. Hence, the measurements M represent the membership functions obtained by the modified fuzzy c-means algorithm:

M I (x, y) = μ I (x, y) (8)

where μ I (x, y) represents the membership degree of pixel p at location (x, y) to the class C according to the image I .

Once the measurements of the information’s coming from the three primitive colors (three information sources) are estimated, their combination is performed using the fuzzy combination rules that gather the T-Norms operators (fuzzy intersection) and the T-Conorms operators (fuzzy union), see Table 1.

After calculating the combination membership degrees for the three images, the decisional procedure for classification purpose consists in choosing one of the most likely hypotheses C , using:

I(x, y) ∈ C ifμ I(x, y) = max{μ I(x, y) ; 1 ≤ k ≤ c} (9)

where 푐 represents the number of classes, and μ I(x, y) represents the membership degree to the class C after the combination step. The proposed method can be described by a flowchart given in Figure 2.



FIGURE 2: FLOWCHART OF THE PROPOSED METHOD.

.

III. EXPERIMENTAL RESULTS AND DISCUSSION In this section, a large variety of color images is employed in our experiments, see Figure 3.

Results obtained on color images are presented in Figs. 4-6. In all the following examples, the decision has been mode using the criterion of maximum membership functions.

In order to evaluate the performance of the proposed method on cancer cells images which is a challenging problem in this field, the segmentation results of the datasets are reported. Consequently, color textured images are developed and used for the numerical evaluation purpose.

Original image Put the image in the HSI color space

Calculate the statistical features of each component image

Compute the 4 significant statistical features extracted from each

component image

Determination of the representative

statistical features

Applied the modified FCM algorithm to determine the membership degrees of

pixels covered the three component images (H, S and I)

Fuzzy Information Modeling

Calculate the T-norms or T-conorms of the membership degrees

Decision

Combination of the three information

sources

Final segmentation results



www.elvedit.com 128

FIGURE 3: DATA SET USED IN THE EXPERIMENT. TWELVE WERE SELECTED FOR A COMPARISON STUDY. THE PATTERNS ARE NUMBERED FROM 1 THROUGH 12, STARTING AT THE UPPER LEFT-HAND CORNER.

The last three images, shown in Figure 3, represents the synthetic images which contains two areas and can be considered as piecewise constant in most of its areas. The other images represented in Figure 3, show real medical cells images, obtained by a himi-histochimy coloring in the Cancer Service previously cited.

Figure 4 shows an example of the fuzzy c-means technique applied to the R, G and B component, respectively. In this case, we can find that the regions are recognized for example in red component but are not identified by the green and blue components. This shows that the RGB space has a strong correlation of its three components, and hence, the use of a single information source leads to bad results.

Comparing the results, we can find that the cells are much better segmented in Fig.4(b) than those in Fig.4 (c) and Fig.4 (d). Also, the first resulting images contain some holes and missing features in one of the cells, which do not exist in the other resulting images. This demonstrates the necessity of using the fusion process.



(A) (B)

(C) (D)

FIGURE 4: SEGMENTATION RESULTS ON A COLOR IMAGE, (A) ORIGINAL IMAGE (256X256X3) WITH GRAY LEVEL SPREAD ON THE RANGE [0,255]. (B) RED RESULTING IMAGE BY FCM METHOD. (C) GREEN RESULTING IMAGE BY

FCM METHOD. (D) BLUE RESULTING IMAGE BY FCM METHOD. Figure 5 displays some examples of segmentations obtained by our algorithm, compared with

other methods [4] [14] [15]. These include Ben Chaabane et al. [4], Vannoorenberghe et al. [14], and Zhu et al. [15]. The segmentation results are shown in Figures 5 and 6.

Figures 5(B), (C), (D) and (E) show the final segmentation results obtained from the DDS algorithm [14], the FCMDS algorithm [15], the HHDS algorithm [4] and our MFCMFF algorithm, respectively. In fact, the partition resulting by the DDS, the FCMDS and the HHDS algorithms are less accurate, and the partition resulting by these methods is not satisfactory either. The performance of the proposed method is quite acceptable.

To evaluate the performance of the proposed segmentation algorithm, its accuracy was recorded. Regarding the accuracy, Table 2 lists the segmentation sensitivity of the different methods for the dataset used in the experiment.



www.elvedit.com 130

TABLE 2. SEGMENTATION SENSITIVITY FROM DDS, FCMDS, HHDS AND MFCMFF FOR THE DATA SET SHOWN IN FIGURE 3.

DDS FCMDS HHDS MFCMFF

(proposed method)

Sensitivity segmentation (%) Image 1 0.9525 0.9741 0.9730 0.9860

Image 2 0.9260 0.9453 0.9782 0.9846

Image 3 0.9491 0.9465 0.9476 0.9879

Image 4 0.9595 0.9604 0.9658 0.9729

Image 5 0.9293 0.9697 0.9706 0.9828

Image 6 0.9565 0.9567 0.9637 0.9702

Image 7 0.9542 0.9550 0.9606 0.9634

Image 8 0.9685 0.9687 0.9798 0.9883

Image 9 0.9492 0.9747 0.9699 0.9714

Image 10 0.9105 0.9350 0.9387 0.9792

Image 11 0.9927 0.9925 0.9972 0.9983

Image 12 0.9765 0.9861 0.9769 0.9894

The segmentation sensitivity [22] [23] is computed using:

푆푒푛(%) = (×

) × 100 (10)

where Sen(%), Npcc, N M are, respectively, the segmentation sensitivity(%), number of correctly classified pixels, and dimension of the image.

Also, the segmentation results by applying the proposed method to textured images are presented. Fig. 6(A) gives the original image. Fig. 6(B) represents the (N × M) original image where a "Saltandpeppers" noise of D density was added. This affects approximately (D × N × M) pixels.

The results obtained by the the DDS, the FCMDS and the HHDS methods are given in Figs. 6(B), 6(C) and 6(D), respectively. Fig. 6(E) shows the segmentation result obtained by our proposed method. Fig. 6(F) shows the reference segmented image.



(A) (B) (C)

(D) (E) (F)

FIGURE 5: COMPARISON OF THE PROPOSED SEGMENTATION METHOD WITH OTHER EXISTING METHODS ON A COMPLEX MEDICAL IMAGE (2 CLASSES, VARIOUS CELLS). (A) ORIGINAL IMAGE (256 × 256 × 3): COLOR MEDICAL IMAGE WITH RGB DESCRIPTION, (B) SEGMENTATION BASED ON DDS METHOD (C) SEGMENTATION BASED ON FCMDS METHOD, (D) SEGMENTATION BASED ON HHDS METHOD, (E) SEGMENTATION BASED ON MODIFIED FUZZY C-MEANS (MFCM) AND FUZZY FUSION (FF), (OUR METHOD), (F) REFERENCE SEGMENTED IMAGE.

(A) (B) (C)

(D) (E) (F)

Figure 6. Comparison of the proposed segmentation method with other existing methods on a textured image (2 classes), (a) original image with RGB representation (256x256x3), (b) original image disturbed with a “salt and pepper” noise, (c) segmentation based on DDS method (d) segmentation based on FCMDS method, (e) segmentation based on HHDS method, (f) segmentation based on Modified Fuzzy C-means (MFCM) and Fuzzy Fusion (FF), (our method).



www.elvedit.com 132

It can be seen that the proposed algorithm produces very satisfactory results for a large variety of images and with the same set of parameters, compared with these other segmentation techniques under the same conditions. In fact, from table 2, one can observe in Figures 6(C), 6(D), and 6(E) that 2.35%, 1.39% and 2.31% of pixels were incorrectly segmented for the DDS, FCMDS and HHDS methods, respectively. Indeed, only 1.16% of pixels were incorrectly segmented in Figure 8(F). Accordingly, the experimental results indicate that the proposed method can provide more accurate results, see Fig. 6(F).

IV. CONCLUSION A new method for segmenting the color images by using a modified fuzzy c-means algorithm

and data fusion techniques has been proposed in this paper. It has been demonstrated that the proposed fusion method with automatic information modeling based on fuzzy c-means and statistical features improves the classification accuracy.

In addition, this new algorithm is less sensitive to the noise presented in the images. It has also been demonstrated that the proposed algorithm can be used successfully for modeling the input information and for segmenting the color images. A quantitative comparison of color images segmentation techniques shows great potential on our novel method. The obtained results demonstrated the significant improved performance in color images segmentation.

Classification accuracies for different existing techniques for various cases considered here also show that the proposed fusion model is consistent. This methodology can also be extended for combining data from dissimilar sources by defining suitable membership functions.

ACKNOWLEDGMENT The authors would like to thank the Editor-in-Chief of IJCIT journal and the anonymous

reviewers. They also address their thanks to Professor Khaled Ben Romdhane, the head of the cancer service of Hospital Salah Azaiez Bab Saadoun, Tunis, Tunisia for providing the diverse images and for his helpful collaboration.

REFERENCES

[1] Kwon MJ, Han YJ, Shin IH, Park HW (2003). Hierarchical fuzzy segmentation of brain MR images. Int. J. Imag. Syst. Technol., 13(2)1, 15-125.

[2] Navon E, Miller O, Averbuch A (2005). Colour image segmentation based on adaptive local thresholds. Image Vision Comput. 23(1), 69-85.

[3] S. Ben Chaabane, M. Sayadi, F. Fnaiech, and E. Brassart (2009). Dempster-Shafer evidence theory for image segmentation: application in cells images. International Journal of Signal Processing, 5(1).

[4] S. Ben Chaabane, M. Sayadi, F. Fnaiech, and E. Brassart (2010). Colour Image Segmentation Using Homogeneity Method and Data Fusion Techniques. EURASIP Journal on Advances in Signal Processing.



[5] Meenavathi MB, Rajesh K (2008). Volterra Filter for Colour Image Segmentation. World Academy of Science. Eng. Technol., 35: 209-214.

[6] N. Ohta (1977). Correspondance between ceilab and cieluv color differences. 2(4), 178-182.

[7] G. Wyszecki and W. S. Stiles, “Color Science: Concepts and Methods, Quantitative Data and formulae”, John Wiley and Sons, second edition, 1982.

[8] F. Kurugollu, B. Sankur and A. E. Harmanci (2001). Color image segmentation using histogram multithresholding and fusion. 19(2001), 915-958.

[9] R. R. Yager (1999). Class of fuzzy measures generated from a Dempster-Shafer belief structure. International Journal of Intelligent Systems, 14(12), 1239–1247.

[10] J. C. Bezdek, “Pattern recognition with fuzzy objective function algorithms,” Plenum Press, New York, 1981.

[11] R. Duda and P. Hart, “Pattern Classification and Scene Analysis”, New York, Wiley, 1973.

[12] Liew AW, Leung SH, Lau WH (2000). Fuzzy image clustering incorporating spatial continuity. IEE Proc. Visual Image Signal Process, 147(2): 185-192.

[13] Ahmed MN, Yamany SM, Mohamed N, Farag AA, Moriarty T (2002). A modified fuzzy c-means algorithm for bias field estimation and segmentation of MRI data. IEEE Trans. Med. Imag. 21(3), 193-199.

[14] P. Vannoorenberghe, O. Colot and D. D. Brucq (1999). Colour image segmentation using Dempster-Shafer’s theory. Pro. ICIP’99. 300-304.

[15] Zhu YM, Dupuis O, Rombaut M (2002). Automatic determination of mass functions in Dempster-Shafer theory using fuzzy c-means and spatial neighborhood information for image segmentation. Opt. Eng. 41(4), 760-770.

[16] R. Bradley (2007). A unified Bayesian decision theory. Theory and Decision. 63(3), 233–263.

[17] Bloch and H. Maitre (1997). Fusion of image information under imprecision. in Aggregation and Fusion of Imperfect Information, B. Bouchon-Meunier, Ed., Series studies in fuzziness, Physical Verlag. 189–213.

[18] Lucas and B. N. Araabi (1999). Generalization of the Dempster-Shafer theory: a fuzzy-valued measure. IEEE Transactions on Fuzzy Systems. 7(3), 255–270.

[19] Dubois and H. Prade (2003). Possibility theory and its applications: a retrospective and prospective view. in Proceedings of the IEEE International Conference on Fuzzy Systems. 1(5).

[20] P. Dempster (1967). Upper and lower probabilities induced by multivalued mapping. Annals of Mathematical Statistics. 38, 325–339.

[21] S. deveughale, B. Dubuisson (1994). Adaptability and Possibilistic Combination: Application to Multi-cameras Vision. Traitement du Signal. 11(1994), 563-568.

[22] Grau, A. U. J. Mewes, M. Alcaniz, R. Kikinis, and S. K. Warfield (2004). Improved watershed transform for medical image segmentation using prior information,” IEEE Transactions on Medical Imaging. 23(4), 447–458.



www.elvedit.com 134

[23] R. O. Duda, P. E. Hart, and D. G. Sork (2000). Pattern Classification, Wiley-Interscience, New York, NY, USA.


Rafika Harrabi born in 1981 in Kairouan (Tunisia), she received the BSc degree in Electrical Engineering and the DEA degree in Automatic and Signal Processing from the High school of sciences and techniques of Tunis, respectively in 2007 and 2009. Currently, she is in the last preparation year of its PhD. Its research interests are focused on signal Processing, image processing, classification, segmentation and data fusion.

Ezzedine Ben Braiek obtained his HDR on 2008 in Electrical engineering from ENSET Tunisia. He is, presently, professor in the department of electrical engineering at the technical university ESSTT and manager of the research group on vision and image processing at the CEREP His fields of interest include automatics, electronics, control, computer vision, image processing and its application in handwritten data recognition.


Model of Brand Building and Enhancement by Electronic Marketing Tools: Practical

Implication

Authors





Virginija Jurkutė Institute of Digital Technologies, Faculty of Social Technologies, Mykolas Romeris University


Abstract

Changing customers buying habits and tendency of growing number of purchasers in the internet force companies to move their business or part of its processes to the electronic environment, and it cause the need to evolve a marketing strategy and its implementation measures. Electronic marketing recently forces this field specialists to look for a new ways to satisfy all customer needs and expectations, which often are associated with intangible attributes, such as brand. However, the lack of e-business managers focus on brand meaning to the company is obvious. Quite often it is limited only up to advertisement. But this is only one of the possible electronic marketing tools used in brand building and enhancement process. Still there is a lack of detailed analysis of electronic marketing tools used in brand building and enhancement in electronic environment. The goal of this paper was to design the model of brand building and enhancement by electronic marketing tools. It was done via analysis of brand building peculiarities in electronic environment and strategic brand building and enhancement process. Suggested model helps to understand the electronic marketing tools position, objectives and functions in brand building and enhancement context. The theoretical model was made to implement cyclic brand building and enhancement by electronic marketing tools process. The model consists of six stages – (1) brand idea sources analysis and brand idea identification, (2) brand components selection, (3) factors influencing evaluation of brand in cyberspace, (4) electronic marketing tools selection, (5) brand experience creation, (6) brand review, development and enhancement. This constant brand adoption would create positivecustomers brand experience and enhance the value of brand. Research on the model of brand building andenhancement by electronic marketing tools implementation and possibilities have been conducted.Quantitative research method had been applied by questioning various electronic business and electronicmarketing employers and employees. Results of the research define the practical implication of suggestedmodel.

Key Words Brand, Brand components, Electronic marketing tools, Brand experience.


Copyright © 2014 Helvetic Editions LTD - All Rights Reserved www.elvedit.com

136

I. INTRODUCTION In the current marketing research the brand is identified as one of the major companies,

organizations exclusivity terms, equivalent in value to even the material property and has a wide range of functions. However, there is still no unified description of the brand acceptable to representatives of all areas - marketing, public relations professionals, business executives, lawyers [18]. Brand is ambiguously evaluated in cyberspace questioning its meaning, purpose and impact on the success of the company's activities [42]. Business in cyberspace already exists for several decades, but this space together with the users’ needs is constantly changing. Requirements for brand are also changing. There is no longer pushing method, such monologue is insufficient to assure the need for interaction [15]. Consumers want to talk about themselves and brand, share experiences in brand communities, which they rely on more than official representatives of the brand [23].

Scientific issue. After transferring business processes into cyberspace, opportunities of integration into the international trade are increasing and business is becoming global. In such a business, the brand becomes one of the main competitive tools. Although this topic is widely discussed internationally (for this a significant influence have annual best global brands elections), but still there is a lack of e-business enterprise managers focus on brand value to the company matters. It is usually limited to products sale or services advertisement, but this is just one of the marketing tools in cyberspace.

Object of the research. E-marketing tools for brand building and enhancement process.

Purpose – to conduct the peculiarities of brand building in cyberspace and design the model of brand building and enhancement by electronic marketing tools.

There have been set the following objectives for the above-mentioned purpose to be achieved: 1. Following a strategic brand building and enhancement analysis to determine the

sources of brand ideas and components implementing brand essence; 2. To set influencing factors of brand in cyberspace; 3. To design the model of brand building and enhancement by electronic marketing

tools; 4. To conduct research on brand building and enhancement by electronic marketing

tools in Lithuania.

Methodology – the paper relies on the theoretical-systematic, conceptual comparative and phenomenological analysis, meta-analysis of the previous studies as well as the quantitative research method. A method of dynamic modeling is also applied.

Practical significance. Changing consumer buying habits and the growing trend of online buying population encourages companies to move their business or part of it‘s processes to the electronic environment, which is changing also marketing measures. Electronic marketing novelty makes experts in this field to look for new opportunities and ways to meet current and future customer needs and expectations, which are often associated with the intangible nature of properties, such as brand, it‘s image, interaction with the company's image and reputation.

Model of Brand Building and Enhancement by Electronic Marketing Tools: Practical Implication Tadas Limba, Gintarė Gulevičiūtė, Virginija Jurkutė


137

There is still a lack of detailed scientific and practical analysis of the marketing tools, which should exploit the brand building and strengthening process in cyberspace.

II. STRATEGIC BRAND DEVELOPMENT AND BUILDING PROCESS, BRAND IDEA SOURCES, BRAND COMPONENTS

Brand development depends on understanding the concept of brand creating process. If the brand is considered merely as a sign in company, that identifies and separates one company's products or services from other companies, then creation of the brand is limited to the design principles, incorporating graphical and linguistic expressions, the sign color and verbal adjustment to selected brand target audience, there all the tasks are done by design experts [31]. However, if the brand creation purpose is to develop a rich brand image and associations system, it‘s value should be discovered in the properties, the benefits (functional and emotional), values and personality (if the sign is the human) [31].

The concept of a brand in literature is analyzed for many years, but there is still no unified description of the brand acceptable to representatives of all areas - marketing, public relations professionals, business executives, lawyers. Practical brand professionals do not agree with the scientists on the meaning of that term, and provide a different brand concept [17, 18]. Some highlights the legal concept of brand, used as an instrument of goods or services differentiation and protection [31], the others near the rational character of the brand refers to the emotional significance of the tangible and non- tangible value to the enterprise and for consumers [29, 19, 25, 37, 44, 4, 36, 20].

Special attention requires S. Vargo and R. Lusch introduced new approach to the concept of brand - called service - dominant (S -D) logic, according to which all marketing activities are oriented to the services, because marketing is a continuous social and economic process which focuses on those resources, empowering companies to provide a better value proposition than the competitors and that the company can always serve customers better. So the emphasis is moved from the material to non-material: skills, information and knowledge, interactivity and interaction as well as ongoing relationships [44], perceived as a service to consumers. So the brand concept disclosure should be made through the prism of the services. R. Brodie in accordance with this point of view, considered the brand as a service mark, naming it in relationship-building, the intermediary between the subjects, whose brand perceptions create the single brand value (See Figure 1).



138

FIGURE 1. TRIANGLE OF SERVICES BRAND- RELATIONS - VALUE. SOURCE: BRODIE ET AL, 2006, P. 372

Giving the promises through external marketing activities creates the brand and image of the company associated with the values of each of relevant stakeholders. Interactive marketing is dependent on the confidence in the staff and the company. It is acquired through the user's brand experience. Internal brand marketing is oriented to companies’ personnel, which are essential to the success of the brand building process [5, 21]. The company and its staff combines organizational culture, built around common values, motivating employees and providing the opportunity to stand out in a unique style of behavior of staff, contributing to the brand building and strengthening [41]. Therefore, in building brand, its conceptual idea would be not only consumers, businesses, as a business entity, but also the employees values and cultural attitudes to the branding. Needs of these three brand building process parties can be foreseen and implemented through L. De Chernatony proposed a strategic brand building and enhancement process (See Figure 2).

FIGURE 2. STRATEGIC BRAND BUILDING AND ENHANCEMENT PROCESS. SOURCE: DE CHERNATONY, 2001, P.

34.

Brand Vision

Organizational Culture

Brand Objectives

Audit Brandsphere

Brand Essence

Internal Implementation

Brand Resourcing



139

This integrated brand building and enhancement model requires for a strategic enterprise approach. The process starts with setting brand vision, providing long-term intentions and properly orienteering company’s further activities to build the brand. L. De Chernatony offers to look at the potential brand after ten years and to assess whether a brand will still be suitable for this environment as well as to establish a brand, which should be much more than the increase of profitability.

Proposed Nike, The Body Shop and the Federal National Real Estate Agency examples, where brand provides benefits even to the whole world [14]. Vision of the brand establishing brand values becomes the basis of differentiation and specifically directs the company's personnel activities. However, the proposal is to provide not only for general (existing in the market, which must comply with the brand in order to overcome barriers to entry into the market), but also the unique values. The amount of the values should not be too large, because then it will be hard for the company personnel to implement, but also for consumers to remember. Organizational culture can become a competitive brand advantage, if it will not focus on what the user gets, but how he gets it [14]. It implements brand vision of the values and reach customers through brand exposure and consumer beams. Artefacts (visual brand expression, workers' clothing, and other material attributes of the company's image) are the most visible organizational culture and values expressions, enabling to see an overall picture [32]. Organization cultural values are dependent on the underlying assumptions, covering organizational conviction and perceptions rooted in the subconscious of every employee of the organization [32].

Organization's culture and brand vision compatibility (misunderstanding and confrontation points identification and resolution) is necessary in order to reach brand success and it influences brand orientation. Set long-term goals of the brand are broken into short-term goals, preventing employees depart from brand vision. In brand environmental analysis, there are evaluated the assessment of the internal communication, the balance between the organization and it‘s products distributors, customer decision-making process, competitive brands and the political, economic, social and technological environment. Essence of the brand is revealed through the brand promise to consumers: the brand attributes inform consumers about the benefits, providing an emotional satisfaction associated with the values symbolized the personality traits, characteristics [14]. Essence of the brand analysis helps to decide, what kind of positioning and brand personality development ideas should be implemented within the organization and externally through brand components (called- brand resourcing). According to L. De Chernatony, brand components, such as a brand exclusive name and identification of ownership are interrelated and identifies the level of brand’s dependence to the company, and what degree of autonomy is granted to it. Product brand functional advantages, such as reliable and functional properties, aesthetics, reveal brand essence of functionality implementation, followed by a post- purchase service stage, contributing to the functional suitability of the time choice, social or financial risk reduction. Legal protection is necessary to avoid the risk of counterfeiters and provides high-quality information to the decision to purchase. Symbolic features of the component include implementation of the brand values associated with the brand personality [14].



140

Taking into account L. De Chernatony analysis of brand building and enhancement process, it is this adaptation and the graphic display (See Figure 3).

FIGURE 3. SUPPLEMENTED STRATEGIC BRAND BUILDING AND ENHANCEMENT PROCESS. SOURCE: ADAPTED

BY DE CHERNATONY (2001).

This supplemented brand building and enhancement process is the right strategic plan for

brand developers and implementers to help understand brand building stages, determines



141

which tasks should be fulfilled at each stage. The process provides consistent actions, which enable to turn the brand vision into reality, visible to consumers through brand components, implemented by appropriate marketing tools.

III. FACTORS INFLUENCING THE BRAND IN CYBERSPACE The changing environment requires firms the flexibility and ability to properly adapt

marketing tools, which reveal the true essence of the brand [9], the corporate strategic brand building process. The brand in cyberspace poses new challenges. Uncertainty as a result of brand expansion into cyberspace or space specifically for this new brand-building effectiveness appears. It also raises arguments about the brand significance in cyberspace. There can be distinguished two scientific positions about the brand in cyberspace: the first, stating, that the brand is no longer relevant in cyberspace because of searching, price comparison and choice possibilities [42], the second, highlighting the importance of the brand in the cyberspace [6, 15], for which the brand is considered to be a major confidence [26] and customer loyalty [27] incentives. There is also no consensus on the essence of the brand identity in the electronic and traditional spaces [9, 37]. However, the brand essence of both traditional and cyberspace remains the same. The only difference here it is implementation. Also, contrary to the first position, somebody can argue, that the search engine optimization capabilities cannot displace the brand, because it do not satisfy all the needs of the user, as an example is given a phone call, during which the communication is not one of the search term, but long syntactic units, that help understand users' needs. Also examining the benefits of price comparison research results indicate that only 22 percent of all transactions are fully fulfilled, because of consumers’ dissatisfaction on insufficient communication with the seller [15].

In view of the brand essence, the brand in cyberspace is having not users, but consumers, who need relationships of one-to-one, proven through conversion rates, as constantly adapted and updated, easily globalized, helping to implement the price flexibility [2]. It’s successful implementation is largely dependent on the brand value expression through interactivity, user experience, customization and rapid response to customers' needs [7]. The value to consumer ensures his loyalty proven by frequent visits and the price here is only a single-used attraction tool. Due to the increased flow of information, consumers have great selection options [13], but due to the more active everyday life for them it is not enough time to examine all the options. Companies, in order to adapt to the changing environment and customer needs and opportunities, are forced to make a decision on the choice of the direction of the brand. In order to establish themselves in cyberspace mostly preferred option that the brand is intended to establish the electronic website address (domain name) (for example, Walt Disney, Toyota, Virgin airline company, Skype, etc.) [30]. Many of the scientific literature authors examine the company's electronic sites as a major brand in cyberspace development and enhancement tool. Therefore, it is practically becoming a brand name [34] (Amazon, Google, Facebook, etc.). Companies should take full advantage of Web 2.0 capabilities: right presentation of the brand promise (brochure website of companies do not reflect the brand promise), to create dialogue between the market and it’s participants through information quality, raising recognizable presentation and easy page navigation, allow users to contribute to brand values (expressed per page design, user emotions) creation (withdraw from brand control and having only the moderator functions) [9] and take a shared responsibility for failures; mobilizing brand community to the open discussion sharing the experience about the brand [15]. The concept of



142

brand value has not lost its importance in cyberspace [36]. Three main factors influencing brand development in cyberspace - interactivity, user understanding and communication [35]. There is offered a much wider brand in cyberspace factors influencing list, which includes: brand experience, interactivity, customization (personalization), relevance of information, website design (attractiveness and aesthetic considerations), customer service, electronic order fulfillment, brand relationship quality, community, site functionality and consumer revisit and timing of the review [10].

IV. DESIGNING THE MODEL OF BRAND BUILDING AND ENHANCEMENT BY ELECTRONIC MARKETING TOOLS

Essence of the brand in cyberspace remains unchanged from the existing traditional space, the development remains in the list of strategic decisions, and the implementation is adapted to the environment in which the brand exist, as well as selected the most appropriate marketing tools enabling to reveal its true meaning. In enterprise brand becomes the cornerstone of the success factor, added value not only to the company but also for the consumer. Consumer value is revealed through the experience of the brand. Therefore, one of the most important tasks is brand experience, through the implementation of functional and emotional values for the user, creation [32, 14]. Positive consumer brand experience is the key for brand building and enhancement. After implementation this, company can achieve high conversion into sales in cyberspace rates [8].

Brand experience acquired in all brand and consumer contact points, starting with the search (how easy or difficult to find), and ending with the purchase transaction and service after purchase [12]. The accumulated experience allows the consumer to shape the overall picture and share it with other actors in his environment, and thus to contribute to brand building and enhancement. Consumer understanding, appropriate actions at each contact point and used effective marketing tools can help companies to create a brand experience to promote conversion into sales. Electronic marketing tools selection is determined by the company's brand -building and enhancement process, it depends on the brand idea, selected components of the brand, and significant factors to the company’s brand in cyberspace. Therefore, it is necessary to examine the entire brand building and enhancement process context. Brand-building and enhancement by electronic marketing tools model enables to reveal a consistent brand essence implementation process and e-marketing tools position and their functions in that process. Therefore, considering consumer perceptions, corporate culture and employee values synergy [5], online brand influencing factors [10], brand building strategy [37], and strategic brand building process [14], there is proposed model of brand building and enhancement by electronic marketing tools (See Figure 4).



143

FIGURE 4. MODEL OF BRAND BUILDING AND ENHANCEMENT BY ELECTRONIC MARKETING TOOLS. SOURCE: COMPILED BY THE AUTHORS

This model depicts the cyclical brand building and enhancement process. The process begins with the brand conceptual ideas, which sources are consumer perceptions and values, company culture, values and business strategy, also employees perceptions and values. The idea of a brand should include the brand vision, goals, response to environmental assessment and determine it‘s essence (functional and emotional values). A strategic brand building process is to choose the appropriate plan for brand components, implementing the idea of the brand [14]. Selected brand components (name, sign of ownership, legal protection, functional value, emotional value, risk reduction, shorthand notation, service components) are implemented by taking into account brand in cyberspace influencing factors (interactivity, personalization, information about brand relevance, electronic site design, functionality, customer service (including electronic order fulfillment), virtual brand communities, consumer revisits, relationships with customers quality), and choosing the right e-marketing tools. To reveal each component, there is a need to understand, what influences the corresponding component of the brand in cyberspace and what electronic marketing tools should be used:

1. Brand name, sign of ownership and legal protection issues can be covered within the context of an electronic site name (domain name):

a. The domain name as a brand name, has the legal protection from intellectual property offense (but it depends on the country laws) and indicates the purposes for which it was created (.com - commercial organizations, .mil - Military organization, .gov - governmental organizations, etc.). It also indicates geographical location of brand (.lt, .lv, .eu, etc.). It is also used to distinguish one company's products/services from other companies products/services and perform such other brand functions as “setting goods, services, information origin (source)...setting quality assurance“ [30].

Brand Idea

Consumer perceptions,

values Corporate culture

and values The company's

business strategy Employee

perceptions, values

Brand Components

Name Sign of ownership Legal protection Functional value Emotional value Risk reduction

Shorthand notation Service components

Brand factors in cyberspace

Interactivity

Personalization Information

relevance Design

Functionality Customer service

Community Consumer revisits

Relationship quality

E-m

arke

ting

tool

s

Bran

d ex

peri

ence

Brand review



144

b. Brand can act as an electronic search keyword, which can be used to transfer to the company's electronic site [37]. Users use such search engines like Google, Yahoo and MSN. In Lithuania the most popular is Google search engine. In 2012 it was used by 98.35 % of Internet users in Lithuania [24], it is also necessary to note that 9 out of 10 Google users review only the first page of search results, so there is a need for website search engine optimization, which ensures a high position in search engine rankings lists. It is necessary to emphasize the pre-consumer attitude, that the most reliable brands are higher than others, so if unknown brand is placed in a higher position than it is known, it draws more attention of consumers [22]. It also provides consumer with similar associations as if well-known brand.

2. Functional value may be disclosed not only through the products quality received by the user, but also through the online site functionality, ease of use. The measures used have to allow more comfortable access to the relevant products from the user-selected categories and attributes (size, color, price). In analysis of the value of the functional aspect, there is a proposal to evaluate the usability [35] (select the relationship between comfort and design), as well as to evaluate it‘s communication channel and speed (higher bandwidth technologies), the quality of site and it’s relevance. Even if the company uses the site only as a business card, it must provide information, which is constantly updated.

a. Convenience - high quality constantly updated content, user-friendly navigation, which requires the least effort to find relevant information. Website content is grouped to help find information faster. The response time is one of the most important factors of convenience and should not exceed focus retention time (5-8 sec.).

b. Page design - color, line, graphics, forms promote the brand recognition. However, it is necessary to draw attention to the large graphics and animation amounts that dwarf the text to which search engines respond best. Graphics, logos can also be used to represent the brand values, and text - to reveal personality.

3. The emotional value is one of the main criteria for determining consumer loyalty, frequent revisits and positive recommendations. It includes the brand promise and brand values, expressed through website design, reflecting feelings close to the consumer, through the brand community discussing frankly and sharing the brand experience, personalization, and brand and consumer relationship quality:

a. Personalization - the brand customization to each consumer individually (used for user name, provided the individual proposals, the page itself is adapted to the individual needs of the user) can create the illusion of individualism, which provides the consumer with emotional satisfaction. It is only possible with the accurate data about each user. Information systems technology enables to create of users databases, which can be exploited in different aspects by marketing specialist. Such data collection is often used with the tool like customer relationship management information system (CRM) [40], which among other functions, also helps to identify points of contact with consumers.

b. Brand and consumer relationships quality is inseparable from the quality of interactivity, ability to communicate one-to-one principle [37]. Sent messages to the consumers are personalized, also selected the most appropriate time and form. This makes it possible to



145

create two-way dialogue. The basis for such dialogue becomes a user databases. Emailing can help build interactivity [43], but it is necessary to let the user choose, what they want to read, watch not to irritate them with spam. One of the helpful tools could be web feeds (RSS) [41]. The implementation of interactivity also includes such electronic marketing tools as chat rooms, interactive games, viral marketing, ratings, feedback, recommendations promoting measures.

c. Community building. 92 % of internet users visit online communities [1]. Their appeal can be explained by opportunity to replace the real social environment, and partially to satisfy the needs of self-expression. In virtual communities gather loyal customers, who share their experience, advice for newcomers, disseminate information about the brand, which attracts new consumers. This encourages using the net even small business representatives. It is proposed to install forums in order to help even in the service sector, where most users consult each other, to install tasks and events calendar, organize competitions to encourage sharing experiences. For the companies, which do not have capabilities to create a virtual community in its website it is recommended to advertise in other online communities [1]. For the brand community building can be used other social media tools [39], which at the same time are inseparable from interactivity: blogs, content sharing tools, social networking. While it is difficult to determine the effectiveness of these measures in terms of returns on investment, however, it is possible to assess the effectiveness of consumer behavior, by the number of visits, time spent on the site, comments about the brand in various social networks (Facebook, Twitter).

4. Risk reduction component reflects consumer confidence in the brand. Trust is based on the credibility of the brand [11]. Consumers expect transaction security and privacy at the same time, so they trust more the well-known brands [11]. Assessing this aspect, the company must maneuver between the consumers’ privacy (less data collected, more privacy ensured) and essential amount of collected users data to ensure the appropriate transaction. Close to the risk reduction component it is important to mention orders fulfilment dimension, that links the real space with electronic and combines interactive and traditional/physical experience. Here, the key roles are played by the order fulfillment time and order accuracy, as described in the corresponding website - the consumer gets what he expects [11]. Order fulfillment element by it’s function and meaning is inseparable from the customer service element, and further is considered as a part of customer service.

5. Shorthand notation may be associated with consumer revisits to website. Based on the consumers decision -making process and gained good experience, users, even after receiving a large amount of information about the brands, prefer well-known brands about which they heard or even seen them [23]. Most often they use electronic search sites such as Google or Yahoo, selecting brand as a keyword. If users have experience with the brand and are loyal to it, they return to the website, reducing the length of time of purchase. However, it can encouraged the use of online advertising tools in the leading company's website - banners, specialized advertising, media / video ads, pop-up windows. Also to ensure consumers frequent revisits it is recommended to use other tools such as e -mail, chat rooms, information updates, special offers and the virtual brand community [37]. To reduce consumers' choice of time affiliated marketing can be exploited too [1]. It should be noted that all of the e- marketing



146

tools enable the company to trace consumer behavior and evaluate the effectiveness of the applied measures.

6. Perhaps one of the most important components of the brand - customer service (service component) [7]. It is a decisive factor in the success or failure, attribute of loyalty and business quality. It‘s importance comes from the consumers' perceptions of service quality dimensions, covering the electronic design of the website (helpful and friendly), reliability (the correct execution of orders, deliveries, personal information), the response to the user's needs, trust and even personalization (personalized attention to each customer, personal thanks and an opportunity to submit questions and comments) [33]. Many of these aspects of the options were discussed above. In order to choose the e- marketing tools enable to respond to the such users’ needs, there can be identified intuitive measures, selected according to the collected data about the user behavior in the website (click), comments and recommendations content, evaluation and opinions expressed in social networks, user requirements for direct submission tools such as question/answer fields, the reaction to the user's e-mails and queries. These measures require the company staff work quality, their confidence in the company's brand and the ability to communicate messages, provided in strategic decisions, and the implementation of the brand promise. However, this is not possible without an effective company internal marketing, where information technology and e- marketing tools are invoked [5].

Summarizing brand building and enhancement with electronic marketing tools and the process of it‘s implementation in reality, it should be noted that the brand requires a strategic corporate approach to it. The creation and strengthening should be a priority in the list of strategic decisions. The process must begin with the brand idea, determined by consumers, businesses and employees perceptions and selection of idea essence implementing components. Each component should be revealed in the context of the online brand influencing factors and electronic marketing tools: brand function value - delivered through the electronic functionality of the site (the proper relationship between E- site convenience and design); the name, sign of ownership and legal protection are dealt with in the context of electronic site name (domain name); given the emotional value through personalization, relationship quality, interactivity, e-marketing tools implementing community mobilization; risk reduction component is inseparable from the transactional and data security solutions; shorthand notation is evaluated in the context of consumers revisits through the web search engines, online ads, transferring to corporate websites. All brand components, implemented within the brand in cyberspace factors and electronic marketing tools, will enable to create a positive experience to the consumer. The idea of a brand must be reflected in the brand experience as a result of target effects. All this is possible only having a clear understanding of process of brand building and enhancement by electronic marketing tools, it‘s stages and actions on each of them.



147

V. RESEARCH ON MODEL OF BRAND BUILDING AND ENHANCEMENT BY ELECTRONIC MARKETING TOOLS

A. Research Methodology

Quantitative research method had been applied by questioning various Lithuanian electronic business and electronic marketing employers and employees, whose work relates to brand building and enhancement by electronic marketing tools. There is a lack of scientific evaluation in Lithuania, which could present the brand influence on consumer decision-making process, deal with the brand in cyberspace and the process of creating effective marketing and brand building experience in cyberspace. The vacuum of information on these aspects encouraged to carry out a research.

The aim of the research – is to evaluate the influence of e-marketing tools used for the brand building and enhancement on brand experience and conversion into sales in Lithuania electronic business enterprises. Research was carried out using a questionnaire. The questionnaire consists of questions in accordance with the principles of the questionnaire formation. Respondents were aware of the purpose, relevance of problem. There has been written, that the form is anonymous and the data will be used generally. Questionnaire identifies the key explanations and instructions on how to fill it.

The respondents were selected from Lithuanian e-business. Respondents were selected on the basis of the Department of Statistics of Lithuania. In 2011 (the last data) there were registered 315 e-business companies. In order to determine the sample of a population Panijoto formula had been used [38]:

n = 1/( Δ² + 1/N);

n – sample size (number of respondents to be interviewed);

Δ – permitted error of sample (used to maintain the reliability of the size - 5 per cent);

N – general set, in this case, companies number, which sales of goods and services are carried out in cyberspace.

n = 1/(0,05 ² + 1/315) = 176 respondents

All the questionnaires were completed consistently and correctly, all of their data is seen as credible to identify trends and shape the conclusions. The empirical study is observational, allowing a preliminary assessment of the current situation and development of forecasts.

B. Research Data Analysis

Based on the scientific literature, in order to reach brand success, implemented through brand building and enhancement process, it was found out, that the first emerging brand idea is born according to the organization's business strategy, culture and values cherished close to staff perceptions and values of the brand, dealing with consumer perceptions and values and their decision-making process (decision to buy or not to buy a particular brand). Therefore, the



148

research aimed to fully disclose companies’ position, determine what the sources of the brand ideas are. There had been found out, that many respondents, creating brand idea, take into account organization's business strategy (see Figure 4).

FIGURE 4. INFLUENCE OF ORGANIZATION ACTIVITIES STRATEGY TO CONCEPTUAL IDEA OF THE BRAND

However, for a substantial number of respondents, the organization's strategy remains behind the development of the conceptual idea of the brand – 26% of respondents think, that organization activities strategy do not influence at all the conceptual idea of the brand, 4% think, that do not influence and 18%, that neither influence nor not. Such a position of the companies may result an incomplete brand idea and it could fail, when there will be a change in the business strategy that will change the position of the brand.

The decisive factor of brand value - consumer revisits. It is important the same as the electronic functionality of the website and customer relationship factor. In order to evaluate the combination of these factors, respondent were asked to define the most important factors, which influence brand in cyberspace (see Figure 5).

FIGURE 5. THE MOST IMPORTANT FACTORS, WHICH INFLUENCE BRAND IN CYBERSPACE

4%

4%18%

48%

26%

Do not influence at all Do not influence

Neither influence nor not Influence

Fully influence

73.91%

65.22%

73.91%

91.30%

78.26%82.61%

56.52%

78.26%

82.61%

Interactivity

Personalization

Information relevancy

Design of website

Funcionality of websiteCustomer service

Virtual brand communities

Consumer revisits

Customer relationship quality



149

The most important factors, influencing the brand in cyberspace and creating value, respondents consider electronic website design, customer service and customer relationship quality. The least support received virtual brand community. It can therefore be concluded, that all the surveyed firms in developing and strengthening the brands, evaluate these factors and plan their marketing activities and measures with regard to this.

Further examining the customer brand experience in Lithuanian e-business there was seeking to find out, whether it has an impact on sales conversions in cyberspace (see Figure 6).

FIGURE 6. INFLUENCE OF CUSTOMER BRAND EXPERIENCE TO CONVERSION OF SALES

22% of respondents stated, that the user brand experience completely influences, and 57%, that influences conversion to sales in cyberspace. This data is seen as justified, because the company has a unique opportunity in cyberspace - to pinpoint conversion rates and effective e-marketing tools correlation (if applied tool increased or decreased conversion into sales rates).

The analysis of brand-building and enhancement process of Lithuanian e-business case was seeking to find out, what marketing tools should be used in order to implement all of the brand and its value influencing factors, so respondents were asked to indicate, which e-marketing tools are used by their companies in brand building and enhancement process (see Figure 7).

4%

0%

17%

57%

22%

Do not influence at all Do not influence

Neither influence nor not Influence

Fully influence



150

FIGURE 7. ELECTRONIC MARKETING TOOLS, USED IN PROCESS OF BRAND BUILDING AND ENHANCEMENT

All respondents as the most important electronic marketing tool in implementing brand building and enhancement process define the name of website. 88% of respondents use advertising on search engines in order to build brand and strengthen it, 75% of respondents use e-mail, social networks. Least likely to use are these marketing tools: pop-up windows - 13%, viral marketing, chat rooms, interactive games - 25%.

Although the empirical study is observational and the data obtained cannot fully justify the findings of Lithuania e-business situation in area of the brand, but based on the obtained hypothetical study results, there can be confirmed, that Lithuanian e-marketing techniques, used to process of brand building and enhancement, create a positive consumer brand experience, which leads to the conversion of sales in cyberspace rates. This situation may change in the use of inappropriate e-business marketing tool. Then the brand experience will be negative and will reduce the conversion in cyberspace results.

VI. CONCLUSION AND RECOMMENDATIONS 1. The brand in cyberspace essence and value remains the same as traditional, but differ

in their implementation. It’s idea comes from the organization's strategy, culture and cherished values, compatible with the perceptions and values of employees, brand consumer



151

perceptions and values. Essence of the brand in cyberspace is implemented through the following components of the brand: the brand name, sign of ownership, legal protection, the functional and emotional value, risk reduction, shorthand notation, services. The brand is determined by such factors as: interactivity, personalization, information about the relevance of the brand, site design and functionality, customer service, the virtual brand communities, consumer revisits, customer relationship quality.

2. There are three main brand in cyberspace development influencing factors - interactivity, user understanding and communication [35]. Also there is offered a much wider brand in cyberspace factors influencing list, which includes: brand experience, interactivity, customization (personalization), relevance of information, page design (attractive and aesthetic considerations), customer service, electronic order fulfillment, brand relationship quality, community, site functionality and user reversibility and timing of the review [10]. Companies should take full advantage of using capabilities of these factors influencing brand in cyberspace.

3. Based on analysis of scientific literature, there was constructed a theoretical model, that can be used for the cyclical brand building and enhancement by electronic marketing tools implementation process. The model consists of 6 main phases - the analysis of brand sources of ideas and brand idea identification, the choice of brand components, determination of the factors influencing brand in cyberspace, choosing means of electronic marketing, brand experience development, brand review, development, enhancement enforcement. This process is not complete, after the implementation of cycle, the process can be updated, creating a positive consumer brand experience, which enhance brand value. The model of brand-building and enhancement by the electronic marketing tools, defined in this article, helps to reveal a consistent brand essence and process by implementing e- marketing tools. This model depicts the cyclical brand development and building process. Model and the implementation of it were verified by empirical research.

4. Research on the model of brand building and enhancement by electronic marketing tools implementation and possibilities have been conducted. Quantitative research method had been applied by questioning various electronic business and electronic marketing employers, employees and their activities relates to brand building and enhancement by electronic marketing tools. Although the empirical study is observational and the data obtained cannot fully justify the findings of e-business situation in area of the brand, but based on the obtained hypothetical study results, there can be confirmed, that e-marketing techniques, used in process of brand building and enhancement, create a positive consumer brand experience, which leads to the increased conversion of sales in cyberspace.

REFERENCES [1] Berkley, H. (2007) Web marketing for small business. Vilnius : Logitema.

[2] Berthon P., Pitt L. F. (1996). The World Wide Web As An Advertising Medium. Journal of Advertising Research, No 36, 43–54.



152

[3] Brodie R. J. (2009). From goods to service branding: An integrative perspective. Marketing Theory, No 9, 107–117.

[4] Brodie R. J., Glynn M. S., Little V. (2006). The Service Brand and The Service Dominant Logic: Missing Fundamental Premise or the Need for Stronger Theory? Marketing Theory, No 6, 363–379.

[5] Brodie R. J., Whittome J. R. M., Brush G. J. (2009). Investigating the service brand: A customer value perspective. Journal of Business Research, No 62, 345–355.

[6] Carpenter P. (2000). EBrands: Building an Internet Business at Breakneck Speed. Harvard: Harvard Business School Press Books.

[7] Chaffey D., Mayer R., Johnston K., Ellis-Chadwick F. (2000). Internet Marketing. Harlow: FT Prentice Hall.

[8] Chaffey D., Ellis-Chadwick F., Mayer R., Johnston K. (2006). Internet Marketing: Strategy, Implementation and Practice, 3rd edition. Pearson Education Limited.

[9] Christodoulides G. (2009). Branding in the post-internet era. Marketing Theory, No 9, 141–144.

[10] Christodoulides G., de Chernatony L. (2004). Dimensionalising on- and offline brands composite equity. Journal of Product & Brand Management, 13(3), 168–179.

[11] Christodoulides G., de Chernatony L., Furrer O., Shiu E., Abimbola T. (2006). Conceptualising and Measuring the Equity of Online Brands. Journal of Marketing Management, No 22, 7–8, 799–825.

[12] Davidavičienė V., Gatautis R., Paliulis N., Petrauskas R. (2009). Electronic Business. Vilnius: VGTU.

[13] Davis S. (2000). Brand asset management. San Francisco: Jassey-Bass.

[14] De Chernatony L. (2001). A model for strategically building brands. Brand management, 9(1), 32–44.

[15] De Chernatony L. (2000). Succeeding with brands on the internet. Journal of Brand Management, 186–195.

[16] De Chernatony L., Dall’Olmo Riley F. (1998). Defining a „brand“: beyond the literature with experts interpretation. Journal of Marketing Management, No 14, 417–443.

[17] De Chernatony L., Dall’Olmo Riley F. (1998). Modelling the components of the brand, European Journal of Marketing, No 32 (11/12), 1074-1090.

[18] De Chernatony L., Dall’Olmo Riley F. (1999). Experts’ Views about Defining Services Brands and the Principles of Services Branding. Journal of Business Research, 46(2), 181–192.

[19] De Chernatony L., McDonald M. (2000). Creating powerful brands. Oxford: Butterworth–Heinemann.

[20] De Chernatony L., McDonald M., Wallace W. (2011). Creating powerful brands. Boston (Mass.): Elsevier: Butterworth-Heinemann.



153

[21] De Chernatony L., Segal-Horn S. (2003). The criteria for successuful services brands. European Journal of Marketing, No 37 (7/8), 1095-1118.

[22] Dou W., Lim K. H., Su C., Zhou N., Cui N. (2010). Brand positioning strategy using search engine marketing. MIS Quarterly, 34(2): 261–279.

[23] Endelman D. C. (2010). Branding in the digital age: You’re Spending Your Money in All the Wrong Places. Harward business review, December.

[24] GemiusRanking. Lithuania, Search engines. Retrieved from http://ranking.lt/en/rankings/search-engines.html

[25] Hoeffler S., Keller K. L. (2003). The marketing advantages of strong brands. Brand Management, 10(6), 421–445.

[26] Hoffman D. L., Novak T., Peralta M. (1999). Building consumer trust in online environments: the case for information privacy. Communications of the ACM, 42(4), 80–85.

[27] Ibeh K. I. N., Luo Y., Dinnie K. (2005). E-branding strategies of internet companies: some preliminary insights from the UK. Journal of Brand Management, 12(5), 355–373.

[28] Jevons C., Gabbott M., de Chernatony L. (2000). Internet noise and its effects on brand management. Retrieved from http://www.brandchannel.com/images/papers/noise_final.pdf

[29] Kapferer J. N. (2012). The new strategic brand management: advanced insights and strategic thinking. London: Kogan Page.

[30] Kiškis M., Petrauskas R., Rotomskis I, Štitilis D. (2006). Law of Informatics and IT law. Vilnius: Mykolas Romeris university.

[31] Kotler P., Amstrong G., Sounders J., Wong V. (2003). Principles of marketing. Kaunas.

[32] Kučinskas V., Paulauskaitė A. (2006). Organizational culture and development in private colleges. 144–165. Retrieved from http://skc.vdu.lt/downloads/zurnalo_arch/amk_2/144_165kucinskas.pdf

[33] Lee G. G., Lin H. F. (2005). Customer perceptions of e-service quality in online shopping. International Journal of Retail & Distribution Management, 33(2), 161–176.

[34] Merz M., Czerwinski D., Amblee, N. (2009). Managing and measuring online brand value Retrieved from http://asbbs.org/files/2009/PDF/A/AmbleeN.pdf

[35] Najmaei A., Sadeghinejad Z. (2009). E-Buyology - a Step towards I-Branding: a Holistic Conceptual Paradigm. Creating Global Economies through Innovation and Knowledge Management: Theory & Practice. 834–843.

[36] Olins W. (2006). Brand. Vilnius.

[37] Rowley J. (2004). Online branding. Online Information Review, 28(2), 131–138.

[38] Sakalauskas, V. (1998). Statistics with STATISTICA. Vilnius.

[39] Shao G. (2009). Understanding the appeal of user-generated media: a uses and gratification perspective. Internet Research, 19(1).



154

[40] Simmons J. (2007). “i-Branding”: developing the internet as a branding tool. Marketing Intelligence & Planning , 25(6), 544–562.

[41] Simmons J., Ahmaad S., Clifton R. (2011). Brand and brand nuilding. Vilnius: Alma litera.

[42] Sinha I. (2000). Cost Transparency: The Net’s Real Threat to Prices and Brands. Harvard Business Review, March–April.

[43] Strauss J., Frost R. (2011). E-Marketing, 6th edition. Boston: Pearson/Prentice Hall.

[44] Vargo S., Lusch R. (2004). Evolving to a New Dominant Logic for Marketing. Journal of Marketing, 68(1), 1–17.

[45] Why is it important to track and improve the position. Retrieved from http://www.pozicijos.lt/page/about


Tadas Limba is a Head of Institute of Digital Technologies at Mykolas Romeris University in Vilnius, Lithuania. He got B. Sc. in Politics from Vilnius University, 1999 and B. Sc. in Law from Mykolas Romeris University, 2010. He got M. Sc. in Public Administration from Mykolas Romeris University, 2001 and M. Sc. in Business Law from Mykolas Romeris University, 2012. Tadas Limba also got his Ph. D. in Management and Administration from Mykolas Romeris University, 2009. Tadas Limba is an Associate Professor from 2010. Tadas Limba has published over 20 articles in Lithuanian and foreign scientific

journals, monograph, textbook, focused on e-government and e-business. His additional areas of research and expertise are – IT law regulation and policy; digital content, digital media, privacy and data protection issues. Tadas Limba is a member of Lithuanian Computer Society since 2007. Since 2013 he is Asia Center Board Member, South Korea's representative at Mykolas Romeris University. He is visiting professor at Zaragoza University in Spain. He plays an active role in international communication and development of joint double degree studies program with South Korea Dongseo University. Tadas Limba made presentations in international and national conferences. Tadas Limba is fluent in English, Spanish and Russian, he is also elementary user of German.

Gintarė Gulevičiūtė was born in Panevėžys, Lithuania in 1989. She got B. Sc. in Public Administration in 2008 and helds M. Sc. in Electronic Business Management from Mykolas Romeris University. Now she is an assistant of Institute of Digital Technologies at Mykolas Romeris University. In 2013 and 2014 she has published some papers – “Peculiarities of E-Government Services Implementation in European Union”; “Holistic Electronic Government Services Integration Model: from Theory to Practice” in Lithuanian and foreign scientific journals. Her areas of interest are e-government, e-business,

business communication and digital contents. Gintarė Gulevičiūtė is the coordinator of Digital Content Academy at Mykolas Romeris University. During her study years she has organized innovative conference “Future business 2013“ at the University. She plays an active role in international communication and development of joint double degree studies program with South Korea Dongseo University. Now she is also a coordinator of Summer School of Communication “Science and Art of Communication” for Chinese students at Mykolas Romeris University.



155

Virginija Jurkutė got B. Sc. in Public Administration at the General Jonas Žemaitis Lithuanian Military Academy in 2007. She got M. Sc. in Electronic Business Management at Institute of Digital Technologies, Faculty of Social Technologies, Mykolas Romeris University, Lithuania in 2013. Her research interests: electronic marketing, branding in electronic environment, electronic commerce, electronic business, public relations. Virginija Jurkutė also serves in Lithuanian armed forces.


A Constraint Programming Approach for Scheduling in a Multi-Project Environment

Authors

Marcin Relich Faculty of Economics and Management/Department of Controlling and Computer Applications in Economics/University of Zielona Gora

[email protected] Gora, 65-216, Poland

Abstract

The paper investigates the use of constraint programming techniques for planning and scheduling in the context of a multi-project environment. Duration and cost of a project activity is specified in the form of discrete α-cuts that enable the connection of distinct and imprecise data, and the implementation of a constraints satisfaction problem with the use of constraint programming. Moreover, the paper presents the impact of a number of α-cuts on project planning and scheduling. A comparison of various variants of project completion takes into account criteria such as time and cost of project and strategy for variable distribution. A declarative form of the description of the decision problem allows its implementation in constraint programming languages and facilitates the development of a decision support system. Optimistic, pessimistic, and several intermediate variants of project scheduling can significantly enhance project managers’ comprehension of time and cost variability and uncertainty.

Key Words

constraint satisfaction problem, decision support system, fuzzy project scheduling, project portfolio, project variants

I. INTRODUCTIONProject management is a complex decision making process involving the rigid project deadline

and budget. The traditional approach to project management is to consider corporate projects as being independent of each other. However, in a multi-project environment the vast majority of projects share resources with other projects and thus the major issue is to find a way of handling

A Constraint Programming Approach for Scheduling in a Multi-Project Environment Marcin Relich


resource scarcity according to the overall strategic direction of the corporation [1]. Sources of uncertainty are wide ranging and have a fundamental effect on projects and project management [2-3]. Companies have to deal with uncertainty in project expected performance in the aspect of their internal and external environments. The external environment concerns the economical, political, technological, social, and ecological issues. In turn, the internal environment includes the project risk factors such as schedule, cost, design, and organisational structure [4].

Uncertainty can be defined in several ways. Essentially, it is lack of information, which may or may not be obtainable [5]. Uncertainty is also linked with risk based on the distinction between aleatory and epistemic uncertainty in the following couplet: uncertainty is immeasurable risk; risk is measurable uncertainty [6-7]. The term risk has different meaning to different people according to their viewpoint, attitudes and experiences. Engineers, designers and contractors view risk from the technological perspective, whereas lenders and developers tend to view it from the economic and financial side [8]. According to Project Management Body of Knowledge, project risk is defined as an uncertain event or condition that, if it occurs, has a positive or a negative effect on a project objective [9].

The CPM (Critical Path Method) and PERT (Program Evaluation and Review Technique) has been widely used for project scheduling, helping managers to guarantee the in time and on budget completion of the project [10]. The hypothesis made in CPM that activity durations are deterministic and known is rarely satisfied in real life where tasks are often uncertain and variable [11-12]. The inherent uncertainty and imprecision in project scheduling has motivated the proposal of several fuzzy set theory based extensions of activity network scheduling techniques. Among these extensions can be found, for instance, resource-constrained fuzzy project-scheduling problem [13], criticality analysis of activity networks with uncertainty in task duration [14], fuzzy repetitive scheduling method [15], fuzzy dependency structure matrix for project scheduling [16], non-delay scheduling [17], potential quality loss cost [18], fuzzy critical chain method for project scheduling under resource constraints and uncertainty [19]. Considerable research effort has been recently focused also on the application of constraint programming frameworks in the context of project scheduling [20-22]. Constraint programming (CP) is qualitatively different from the other programming paradigms, in terms of declarative, object-oriented and concurrent programming. Declarative programming languages base on the idea that programs should be as close as possible to the problem specification and domain [23].

The paper aims at using constraint programming to fuzzy project scheduling and cost evaluation in multi-project environment whose durations and costs are in the imprecise form. The model of project portfolio planning is specified in terms of fuzzy constraints satisfaction problem (CSP), using constraint programming to seek a solution to the problem, and enabling cost analysis at different α-cuts. An α-cut is a crisp set consisting of elements of fuzzy set A which belong to the fuzzy set at least to a degree of α. The proposed methodology is relatively similar to what practitioners are using to generate project cost and cash flows but is considerably more



www.elvedit.com 158

effective and realistic in modelling uncertainty. The proposed decision support system for project portfolio planning allows a decision-maker to obtain a set of project variants and to perform analysis of cost uncertainty at different α-levels.

The remaining sections of this paper are organised as follows: Section 2 presents a problem formulation in terms of fuzzy CSP for project portfolio scheduling, a method of fuzzy project scheduling is shown in Section 3, an illustrative example of the proposed methodology is presented in Section 4, in turn, some concluding remarks are contained in Section 5.

II. PROBLEM FORMULATION The constraint programming environment seems to be particularly well suited to modelling

real-life and day-to-day decision-making processes at an enterprise, including project planning and scheduling. Constraint programming is qualitatively different from the other programming paradigms, in terms of declarative, object-oriented and concurrent programming. Compared to these paradigms, constraint programming is much closer to the ideal of declarative programming: to state what we want without stating how to achieve it [24]. CP is an emergent software technology for a declarative constraints satisfaction problem description and can be considered as a pertinent framework for the development of decision support system software.

Declarative programming languages base on the idea that programs should be as close as possible to the problem specification and domain [23]. In the field of constraint-based scheduling, two strengths emerge: natural and flexible modeling of scheduling problems as CSP and powerful propagation of temporal and resource constraints. Thus, the scheduling problem is modelled as CSP at hand in the required real-life detail and it enables to avoid the classical drawbacks of being forced to discard degrees of freedom and side constraints. Discarding degrees of freedom may result in the elimination of interesting solutions, regardless of the solution method used. Discarding side constraints gives a simplified problem and solving this simplified problem may result in impractical solutions for the original problem [25]. The limitations of imperative languages provide the motivation to develop a reference model of project management in an enterprise and to implement it in declarative languages. The advantage of working with such a model is that users are driven by the system to produce the required results, whilst the manner in which the results are produced is dependent on the preferences of the users [26].

The model formulated in terms of CSP determines a single knowledge base and it enables effective implementation in constraint programming languages, as well as the development of a task-oriented decision support system (DSS) for project portfolio planning. As a result, the problem specification is closer to the original problem, obtaining solutions that are unavailable with imperative programming. The specification of project portfolio scheduling encompasses technical parameters, expert’s experiences and user expectations in the form of a knowledge base, i.e. as a set of variables, their domains, and a set of relations (constraints) that restrict and link variables. In this context, it seems natural to classify some decision problems as CSP. The problem formulation in terms of CSP enables a simplified description of actuality, i.e. a



description encompasses the assumptions of object, implementing therein tasks, and a set of routine queries – the instances of decision problems [27-30].

In a classical form, the structure of the constraints satisfaction problem may be described as follows [20], [28]: CSP = ((V, D), C), where: V – a set of variables, D – a set of discrete domains of variables, C – a set of constraints. In turn, for the imprecise description of variables, the Fuzzy Constraints Satisfaction Problem (FCSP) takes the following form:

)),,~

(( CDVFCSP (1)

where: nvvvV ~,...,~,~~21 – a finite set of n fuzzy variables that are described in the form of fuzzy

number (a finite set of discrete α-cuts); D = {d1, d2, ..., dn} – a set of domains for n fuzzy variables; C = {c1, c2, ..., cm} – a finite set of m constraints limiting and linking decision variables.

Given a set of projects P = {P1, P2, …, PI}, where the project Pi consists of J activities: Pi = {Ai,1, ..., Ai,j, ..., AI,J}. The j-th activity of i-th project that is specified as follows: Ai,j = {si,j, zi,j, ti,j, dpi,j}, where:

si,j – the starting time of the activity Ai,j, i.e., the time counted from the beginning of the time horizon H;

zi,j – the completion time of the activity Ai,j;

ti,j – the duration of the activity Ai,j, si,j < zi,j;

dpi,j – the financial means allocated to the activity Ai,j.

The project Pi is described as an activity-on-node network, where nodes represent the activities and the arcs determine the precedence constraints between activities. According to this, the precedence constraints are as follows:

for the k-th activity which follows the j-th one: si,j + ti,j ≤ si,k;

for the k-th activity which follows other activities: si,j + ti,j ≤ si,k, si,j+1 + ti,j+1 ≤ si,k, ..., si,j+n + ti,j+n ≤ si,k;

for the k-th activity which is followed by other activities: si,k + ti,k ≤ si,j, si,k + ti,k ≤ si,j+1, ..., si,k + ti,k ≤ si,j+n.

CSP can be considered as a knowledge base that is a platform for query formulation as well as for obtaining answers, and it comprises of facts and rules that are characteristic of the system’s properties and the relations between its different parts [4]. As a consequence, a knowledge base



www.elvedit.com 160

facilitates the implementation of a decision support system [31-32]. The distinction of decision variables that are embedded in the knowledge base as an input-output variable permits to formulate the standard routine queries concerning project cost analysis such as: is there a schedule at given α-cut, project deadline, budget, precedence constraints, etc? The method of generation of admissible solutions for the above-described problem is presented in the next section.

III. METHOD OF FUZZY PROJECT SCHEDULING WITH THE USE OF Α-CUTS The perception or estimation of uncertainty is encoded in the initial assignment of fuzzy

activity duration and cost [33]. Several researchers have applied different approaches to fuzzy set theory or probability theory in project flow generation and analysis (e.g. [34-36]). In terms of project management, different α-cuts can be considered as separate risk levels [37]. The risk levels can vary from “none”, “low”, “moderate”, “high” to “very high” as the α-cut moves from 1 towards 0. The difference between the proposed approach and PERT network diagrams concerns the number of variants and the use of integer numbers. PERT assumes only the absolute worst and best scenarios (everything goes worse or better than expected, respectively), whereas the proposed approach includes some possibility levels from 0 to 1 [33].

Imprecise variables determined by convex membership function μ(t) (e.g. a triangular fuzzy number t = <a, b, c>) can be specified as α-cuts. An α-cut is a crisp set consisting of elements belong to the fuzzy set at least to a degree of α (0 < α ≤ 1). An α-cut is a method of defuzzifying a fuzzy set to a crisp set at desired α-levels that correspond to the perceived risk (α = 1 meaning no risk, α = 0– meaning the lowest risk, α = 0+ meaning the highest risk). Additionally, the low (α = 0–) and high (α = 0+) values of every α-cut represent the optimistic and pessimistic outcomes of that risk level. The main objective of fuzzy project scheduling is to apply fuzzy set theory concepts to the scheduling of real world projects where task duration can be specified as fuzzy numbers instead of crisp numbers [27].

The fuzzy project scheduling algorithm requires the specific assumptions. The fuzzy form of the start and completion times can lead to difficulties with the interpretation, if the fuzzy starting time of the activity is greater than the fuzzy completion time. In the order to avoid this situation, it is assumed that the starting time of the activity is in a distinct form, whereas the completion time of the activity can be specified as a fuzzy number. The fuzzy completion time is the sum of the activity start with the fuzzy activity duration (see Figure 1). It is noteworthy that using the presented methodology, the intersection of starting and completion time is impossible and the interpretation is unambiguous.



FIGURE 1: ADDITION OF FUZZY NUMBERS IN TERMS OF DISCRETIZED Α-CUTS

The duration of activity varies for different number of α-cuts, and determining optimistic,

pessimistic and several intermediate variants. In the best case (mint0), the activity starts as early as possible and lasts the minimum duration. In the worst case (maxt0), the activity starts as late as possible and lasts the maximum duration. An example of the duration of an activity three α-levels (mint0, t1, maxt0) is presented in Figure 2.

FIGURE 2: START AND FUZZY COMPLETION TIME OF ACTIVITY Figure 2 shows an activity with starting time of <10, 10, 10>, duration of <5, 6, 7>, and

completion time of <15, 16, 17>. In this example, the duration intervals at α = 0 are mint0 = [10, 15] and maxt0 = [10, 17] and the activity cost is distributed in these intervals. In the optimistic variant, the activity begins as early as possible (10th time unit) and lasts the minimum duration (5 time units), whereas in the pessimistic variant, it lasts the maximum duration (7 time units).

The uncertainties of the duration and cost of an activity are positively correlated, so the minimum (mint0) and maximum (maxt0) cost distribution per unit of time of the j-th activity at the level α depict the best and the worst variant, respectively. A number of variants depend on a number of α-cuts and the shape of a fuzzy number. For instance, if a fuzzy number is triangular and described at 3 α-cuts, then there can be 3 variants (mint0, t1, maxt0) or 5 variants (mint0, mint0.5, t1, maxt0.5, maxt0) of project schedule description. This case is presented on the left and right side of Figure 3, respectively.



www.elvedit.com 162

FIGURE 3: SHAPE OF FUZZY NUMBER FOR 3 AND 5 VARIANTS An example of the use of the presented methodology in constraint programming environment is

presented in the next section.

IV. ILLUSTRATIVE EXAMPLE The example consists of the following parts: the description of the project portfolio, the analysis

of the first admissible solution of the fuzzy scheduling problem for the different variants, the analysis of cumulative cost for these variants, and the impact of strategy of variable distribution on searching the first admissible solution. It is assumed that the time horizon for the project portfolio (P = {P1, P2}) equals 26 months and the budget of the project portfolio is fixed at 300 m.u. The network diagrams of the activities in the project portfolio are shown in Figure 4 and 5.

FIGURE 4: NETWORK DIAGRAM FOR PROJECT P1

FIGURE 5: NETWORK DIAGRAM FOR PROJECT P2 The duration of some activities (A1,7, A1,9, A1,10, A1,11, A1,12, A2,6, A2,7, A2,8, A2,10, A2,11) is specified

in the imprecise form. The sequences of activity duration for the considered projects can be described as follows: T1 = (3, 3, 2, 1, 2, 3, “about 6”, 2, “about 3”, “about 4”, “about 3”, “about 6”), T2 = (4, 3, 4, 3, 2, “about 3”, “about 4”, “about 3”, 2, “about 4”, “about 5”). For instance, the



duration of the activity A1,7 is “about 6”, i.e. the activity can be executed within the time period of 5 to 7 units of time.

Fuzzy project scheduling and cost generation problem can be reduced to the following questions: is there a portfolio schedule (and if yes, what are its parameters) that follows from the given project constraints specified by the duration of activities, the deadline and budget of project portfolio? The answer to the question is connected with the determination of the starting (si,j,α) and completion (zi,j,α) time of project portfolio activities and the allocation of financial means to the activities by different α-level dpi,j,α. The example includes three cases of project scheduling, for optimistic variant (mint0), intermediate variant (t1) and pessimistic variant (maxt0). For the considered project portfolio, the following sequences are sought: S1,α = (s1,1,α, …, s1,12,α), S2,α = (s2,1,α, …, s2,11,α), Z1,α = (z1,1,α, …, z1,12,α), Z2,α = (z2,1,α, …, z2,11,α), Dp1,α = (dp1,1,1, …, dp1,12,α), Dp2,α = (dp2,1,1, …, dp2,11,α), where α equals 0–, 1, 0+.

Figure 6 presents the first admissible solution (project portfolio schedule) for the optimistic variant (α = 0–). The sequences of activity starting and completion time are as follows: S1,0- = (0, 3, 3, 6, 5, 6, 7, 7, 5, 12, 9, 15), S2,0- = (0, 4, 4, 4, 7, 7, 8, 7, 9, 11, 14), Z1,0- = (3, 6, 5, 7, 7, 9, “about 13”, 9, “about 8”, “about 16”, “about 12”, “about 21”), Z2,0- = (4, 7, 8, 7, 9, “about 10”, “about 12”, “about 10”, 11, “about 15”, “about 19”).

FIGURE 6: PROJECT PORTFOLIO SCHEDULE FOR OPTIMISTIC VARIANT



www.elvedit.com 164

The project portfolio schedule for intermediate variant (α = 1) is illustrated in Figure 7. This variant is equivalent to that generated from deterministic analysis. There are the following sequences of activity starting and completion time: S1,1 = (0, 3, 3, 6, 5, 6, 7, 7, 5, 13, 9, 17), S2,1 = (0, 4, 4, 4, 7, 7, 8, 7, 9, 12, 16), Z1,1 = (3, 6, 5, 7, 7, 9, “about 13”, 9, “about 8”, “about 17”, “about 12”, “about 23”), Z2,1 = (4, 7, 8, 7, 9, “about 10”, “about 12”, “about 10”, 11, “about 16”, “about 21”).

FIGURE 7: PROJECT PORTFOLIO SCHEDULE FOR INTERMEDIATE VARIANT The project portfolio schedule for pessimistic variant (α = 0+) is shown in Figure 8. The

sequences of activity starting and completion time are as follows: S1,0+ = (0, 3, 3, 6, 5, 6, 7, 7, 5, 14, 9, 19), S2,0+ = (0, 4, 4, 4, 7, 7, 8, 7, 9, 13, 18), Z1,0+ = (3, 6, 5, 7, 7, 9, “about 13”, 9, “about 8”, “about 18”, “about 12”, “about 25”), Z2,0+ = (4, 7, 8, 7, 9, “about 10”, “about 12”, “about 10”, 11, “about 17”, “about 23”).



FIGURE 8: PROJECT PORTFOLIO SCHEDULE FOR PESSIMISTIC VARIANT The time of project completion for the different variants depends on a number of activities that

are specified in imprecise form and the shape of a fuzzy number (see Figure 6-8). The greater base of a fuzzy number implicates more variants and the longer time of project completion (see Figure 3).

Figure 9 presents three different cost variants for project portfolio (cumulative cost for project P1 and P2). At α = 1, the cash flow (dotted line) is equivalent to that generated from deterministic analysis. At α = 0, there is an optimistic variant below and a pessimistic one above (solid line). In the optimistic variant (mint0), the project portfolio will be completed in 20 months with the total cost of 195 m.u., whereas in the pessimistic variant (maxt0) in 26 months with the total cost of 255 m.u.



www.elvedit.com 166

FIGURE 9: CUMULATIVE COST FOR PROJECT PORTFOLIO The presented S-curves are the basis for analyzing cost variants in project portfolio. In order to

develop this analysis, S-curves can be widespread for other α-levels (e.g. 0, 0.1, 0.2, …, 1) and plotted an S-surface. Figure 10 shows the project S-surfaces for the optimistic and pessimistic variant. The selection of specific possibility levels and time intervals determines the size of the rectangular patches that form the S-surface and consequently the overall plot quality.

FIGURE 10: S-SURFACES FOR OPTIMISTIC AND PESSIMISTIC VARIANT OF PROJECT COST

Compared with conventional 2 dimensional S-curves, the S-surface shows how both uncertainty

levels and time affect the project cost. Thus, the surface steepness in terms of possibility and time provides additional insight about the project cost. At α = 1, the optimistic and pessimistic S-surface intersect each other.

The presented in Figure 6-8 project portfolio schedules concern the first admissible solution.



The sought solutions can be evaluated according to criteria such as strategy of variable distribution, number of nodes, depth of the search tree. Table 1 presents the results for three strategies of variable distribution: First-fail, Naïve, and Split. The example was implemented in the Oz Mozart programming environment and tested on an AMD Turion(tm) II Ultra Dual-Core M600 2.40GHz, RAM 2 GB platform. The results show that the First-fail and Split distribution strategy significantly outperforms the Naïve ones.

TABLE 1: COMPARISON OF STRATEGIES FOR VARIABLE DISTRIBUTION Distribution

strategy Number of

choice nodes

Number of failed nodes

Depth Time [sec]

First-fail 822 807 27 0.312 Naïve 20,226 20,211 31 8.20 Split 840 807 37 0.281

The presented approach allows the decision-maker to consider a wide range of further analyses.

For instance, a risk level for cost variant can be treated as an additional criterion for reducing a set of admissible project alternatives in the feasibility study. Moreover, the obtained schedules and cost variants provide a plan for project portfolio execution and are a basis for further adjustment aimed at fitting to real live execution.

V. CONCLUSIONS Most projects are executed in the presence of uncertainty and are difficult to manage, especially

in a multi-project environment. Concurrent projects are often interrelated due to interdependencies between inputs and outputs and sharing of specialized resources [38]. Hence, a pure deterministic approach for evaluating project time and cost is inadequate. The proposed approach takes into account several elements, such as the distinct and imprecise duration of project activities, cost distribution analysis, including S-curves and S-surfaces. Data specification in the form of α-cuts enables the generation of a set of project variants (schedules) that can significantly enhance project managers’ comprehension of time and cost variability and uncertainty. Moreover, the use of discrete α-cuts facilitates the combination of distinct and imprecise data, and implementation of a constraints satisfaction problem in the constraint programming environment that solves CSP with a significant reduction of the amount of search space. As a result, a task-oriented decision support system has been effectively developed [39]-[40]. This system can support a decision-maker in obtaining project portfolio scheduling and fuzzy project cost generation.

The limitations of existing commercially available tools (e.g. lack of possibility for data specification in an imprecise form, lack of abilities to solve problems defined in multi-project environments) was the motivation to develop a design methodology for a task-oriented decision support system that bases on the constraint programming techniques. The proposed methodology can provide a better perception of risk which is usually obscured in the conventional approach. The number of α-levels can be modified according to the decision-maker’s requirements and can



www.elvedit.com 168

assist project managers to gain deeper insight into the sources and extents of uncertainty, which may in turn lead to the avoidance of troubles during project implementation. In addition, the presented methodology concerns the assessment of financial requirements during project implementation and it can be useful in evaluating alternative variants of project portfolio during the feasibility stage. Moreover, it tends to achieve a balance between complexity of methodology and an intuitive, effective decision support system that is realistic in modeling uncertainty. Finally, its application in performing earned value analysis during project monitoring can also obtain useful results [33]. The subject of future research can focus on developing the proposed approach in the aspect of system thinking and multiculturality in project management, for instance, according to the results presented in [41-45].

REFERENCES [1] Laslo, Z. (2010). Project portfolio management: An integrated method for resource planning and

scheduling to minimize planning/scheduling-dependent expenses. International Journal of Project Management, 28, 609-618.

[2] Atkinson, R., Crawford, L., Ward, S. (2006). Fundamental uncertainties in projects and the scope of project management. International Journal of Project Management, 24, 687-698.

[3] Relich, M. (2012). An evaluation of project completion with application of fuzzy set theory. Management, 16(1), 216-229.

[4] Relich, M., Banaszak, Z. (2011). Reference model of project prototyping problem. Foundations of Management, 3(1), 33-46.

[5] Rowe, W. D. (1994). Understanding uncertainty. Risk Analysis, 14, 743-750.

[6] Hillson, D. (2004). Effective opportunity management for projects – exploiting positive risk. New York: Marcel Dekker.

[7] Olsson, R. (2007). In search of opportunity management: is the risk management process enough? International Journal of Project Management, 25, 745-752.

[8] Baloi, D. & Price, A. D. F. (2003). Modelling global risk factors affecting construction cost performance. International Journal of Project Management, 21, 261-269.

[9] Project Management Institute (2000). A guide to the project management book of knowledge (PMBOK). Newtown Square.

[10] Kastor, A. & Sirakoulis, K. (2009). The effectiveness of resource levelling tools for Resource Constraint Project Scheduling Problem. International Journal of Project Management, 27, 493-500.

[11] Zammori, F. A., Braglia, M., Frosolini, M. (2009). A fuzzy multi-criteria approach for critical path definition. International Journal of Project Management, 27, 278-291.

[12] Carr, V. & Tah, J. H. M. (2001). A fuzzy approach to construction project risk assessment and analysis: construction project risk management system. Advances in Engineering Software, 32, 847-857.



[13] Bonnal, P., Gaurc, K., Lacoste, G. (2004). Where do we stand with fuzzy project scheduling? Journal of Construction Engineering and Management, 130, 114-123.

[14] Fortin, J., Zielinski, P., Dubois, D., Fargier, F. (2010). Criticality analysis of activity networks under interval uncertainty. Journal of Scheduling, 13, 609-627.

[15] Maravas A. & Pantouvakis J. P. (2011). A fuzzy repetitive scheduling method for projects with repeating activities. Journal of Construction Engineering and Management, 137, 561-564.

[16] Shi, Q. & Blomquist, T. (2012). A new approach for project scheduling using fuzzy dependency structure matrix. International Journal of Project Management, 30, 503-510.

[17] Zwikael, O., Cohen, Y., Sadeh, A. (2006). Non-delay scheduling as a managerial approach for managing Project. International Journal of Project Management, 24, 330-336.

[18] Kim, J. Y., Kang, Ch. W., Hwang, I. K. (2012). A practical approach to project scheduling: considering the potential quality loss cost in the time-cost tradeoff problem. International Journal of Project Management, 30, 264-272.

[19] Long, L. D. & Ohsato, A. (2008). Fuzzy critical chain method for project scheduling under resource constraints and uncertainty. International Journal of Project Management, 26, 688-698.

[20] Banaszak Z., Zaremba M., Muszyński W. (2009). Constraint programming for project-driven manufacturing. Int. J. Production Economics, 120, 463-475.

[21] Relich, M. (2011). CP-based decision support for scheduling. Applied Computer Science, 7(1), 7-17.

[22] Relich, M. (2011). Project prototyping with application of CP-based approach. Management, 15(2), 364-377.

[23] Hofstedt, P. & Pepper, P. (2007). Integration of declarative and constraint programming. Theory And Practice of Logic Programming, 7, 93-121.

[24] Van Roy, P., Haridi, S. (2004). Concepts, techniques and models of computer programming. Massachusetts Institute of Technology.

[25] Rossi F., van Beek P., Walsh T. (2006). Handbook of Constraint Programming. Elsevier Science.

[26] Pesic M., & van der Aalst W. M. P. (2006). A declarative approach for flexible business processes management. In Business Process Management Workshops, Lecture Notes in Computer Science, Springer, 4103, 169-180.

[27] Relich, M. (2013). Fuzzy project scheduling using constraint programming. Applied Computer Science, 9(1), 3-16.

[28] Relich, M. (2014). A declarative approach to new product development in the automotive industry. In Environmental Issues in Automotive Industry (pp. 23-45), Springer Berlin Heidelberg.

[29] Sitek, P. & Wikarek, J. (2013). A hybrid method for modeling and solving constrained search problems. In Proceedings of the Federated Conference on Computer Science and Information Systems (pp. 385-392), Krakow, Poland.



www.elvedit.com 170

[30] Sitek, P. & Wikarek, J. (2013). A hybrid approach to supply chain modeling and optimization. In Proceedings of the Federated Conference on Computer Science and Information Systems (pp. 1211-1218), Krakow, Poland.

[31] Relich, M. (2013). Knowledge acquisition for new product development with the use of an ERP database. In Proceedings of the Federated Conference on Computer Science and Information Systems (pp. 1285-1290), Krakow, Poland.

[32] Relich, M., Witkowski, K., Saniuk, S., Sujanova, J. (2014). Material demand forecasting: an ERP system perspective. Applied Mechanics and Materials, 527, 311-314.

[33] Maravas, A. & Pantouvakis, J. P. (2012). Project cash flow analysis in the presence of uncertainty in activity duration and cost. International Journal of Project Management, 30, 374-384.

[34] Barraza, G. A., Back, W. E., Mata, F. (2004). Probabilistic forecasting of project performance using stochastic S curves. Journal of Construction Engineering and Management, 130(1), 25-32.

[35] Yao, J. S., Chen, M. S., Lu, H. F. (2006). A fuzzy stochastic single-period model for cash management. European Journal of Operational Research, 170(1), 72-90.

[36] Cheng, M. Y. & Roy, A. F. V. (2011). Evolutionary fuzzy decision model for cash flow prediction using time dependent support vector machines. International Journal of Project Management, 29, 56-65.

[37] Mon, D. L., Cheng, C. H, Lu, H. C. (1995). Application of fuzzy distribution on project management. Fuzzy Set and Systems, 73(2), 227-234.

[38] Yaghootkar, K. & Gil, N. (2012). The effects of schedule-driven project management in multi-project environments. International Journal of Project Management, 30, 127-140.

[39] Relich, M. (2010). A decision support system for alternative project choice based on fuzzy neural networks. Management and Production Engineering Review, 1(4), 46-54.

[40] Relich, M. (2010). Assessment of task duration in investment projects. Management, 14(2), 136-147.

[41] Peric, M. & Djurkin, J. (2014). Systems thinking and alternative business model for responsible tourist destination. Kybernetes, 43(3/4), 480-496.

[42] Relich, M. (2006). Liquidity management in a small enterprise by the application of reverse approach. Management, 10(2), 129-137.

[43] Caganova, D., Cambal, M., Sujanova, J., Woolliscroft, P. (2012). The Multiculturality Aspects and Human Capital Management within Slovak Industrial Enterprises. In Proceedings of the 4th European Conferences on Intellectual Capital (pp. 106-117), Helsinki, Finland.

[44] Caganova, D., Cambal, M., Weidlichova Luptakova, S. (2010). Intercultural Management – Trend of Contemporary Globalized World. Electronics and Electrical Engineering, 6(102), 51-54.

[45] Kormancova, G. & Theodoulides, L. (2013). The intercultural dimensions of the cultures in transition process in Central and Eastern Europe. In Contemporary Challenges towards Management III (pp. 41-60), Uniwersytet Slaski w Katowicach.




Marcin Relich, PhD, is an Assistant Professor at the Faculty of Economics and Management, Department of Controlling and Computer Applications in Economics, University of Zielona Gora, Poland. His current research interests include various aspects of information systems development and management, enterprise application integration, business process improvement, and knowledge management. His numerous publication activities are closely connected with artificial intelligence methods and tools, project

management, and innovation. His previous research interests were connected with decision support systems and early warning systems in enterprises, operations research, and production engineering.


Anti-Crisis Management Tools for Capitalist Economy

Authors

Alexander A. Antonov Research Centre of Information Technologies “TELAN Electronics”

[email protected] Kiev, 03142, Ukraine

Abstract

It is stated that the reason for ineffective economic crisis fighting is misunderstanding of processes prevailing in the economy, as well as lack of mathematical apparatus necessary for their description. This mathematical description was found using the analogy approach – it turned out that the electric circuit theory is the ‘white box’ for the ‘black box’ of economics. Mathematical description of processes in the electric circuit theory allowed describing mathematically the behaviour of the market participants and deriving the differential equation for the ‘goods-money-goods’ process. Its analysis enabled to reveal structural defects of the capitalist economy – its nonlinearity and destructive influence of the human factor. These defects cause economic crises due to instability of processes prevailing in the capitalist economy. Tools for management of the capitalist economy in the form of business-interfaces and new global computer network TV•net, which will provide for its crisis-proof development, are suggested.

Key Words

Computer network, crisis-proof economy, economic reforms, super-intelligence, subconscious thinking.

I. INTRODUCTIONThe first economic crisis happened in England in 1825 [1]. Since that time the global economy

is prone to this inevitably repeated disaster.

At first, after every next economic crisis, some of the prominent academic economists insisted that they know how to grapple with it. They were given the opportunity; however, each subsequent crisis disproved theoretical assumption.

At present, no one insists any more that they know how to grapple with economic crises.

Anti-Crisis Management Tools for Capitalist Economy Alexander A. Antonov


II. CAUSES OF CRISES IN THE CAPITALIST ECONOMY Nevertheless, economic crises must be combatted. There must be a solution to this problem,

because there is at least one known variant of a crisis-proof economy. It was even implemented in practice. It is the socialist economy. However, it existed only in the totalitarian regime countries that have certain well-known drawbacks.

This is why, over the past decades, many attempts were taken to create a hybrid capitalist-socialist economy, and the solutions varied from country to country. However, these attempts were futile. Economic crises still occur.

A. Lack of capitalist economic theory

It is not surprising, because the global economic science is groping with the problem in the dark, being unable to describe economic processes mathematically. Over almost 200 years since the first economic crisis, economics has not realized that first and foremost it needs this mathematical knowledge.

It is impossible to imagine that without mastering not only theoretical radio electronics, but mathematics, as well, anyone would be able to assemble a TV set or a computer. As for economics, quite recently prominent scientists insisted that economics could do without mathematics.

However, economic scholars do not realize it and are making no headway by letting no young ideas, for instance, econophysics, into their science. However, this is the shortcoming of many other sciences, including physics, where monopolies referred to as ‘ scientific schools ’ successfully suppress scientific dissent [2].

As for the mathematical tools currently used in economics, they are based mostly on mathematical statistics and graphical methods of solving algebraic equations (e.g., the supply and demand curves). However, these mathematical tools allow defining only states. As a result, the obtained low-factor information about the significantly multi-factor economic processes is scarce and does not enable to understand what is going on in the economy. Therefore, economists have to cure the economy, to put it figuratively, using the mean temperature of all patients in a hospital. No doctor would ever do that.

This is why some scientists [3] doubt whether it is possible to develop economics as an exact science. Opinions were expressed that, apparently, a large variety of economic phenomena cannot be accounted for based on a limited number of fundamental laws. It was even suggested to substitute the principle of an integral economic theory for the principle of coexistence of competing concepts [4].

However, the trouble is that competing concepts cannot be developed without understanding the processes prevailing in the economy. Eventually, despite the attempts to use more and more sophisticated mathematical tools in economics in the recent decades [5], contemporary economic theory is not able to understand these processes and does not meet the criteria of the exact sciences [6].



www.elvedit.com 174

B. Structural defects of the capitalist economy

Use of the mathematical apparatus discussed below, which is able to describe economic processes, allowed revealing a number of structural defects in the modern capitalist economy, which, in fact, cause economic crises.

1. Differential equation of the ‘goods-money-goods’ process: In the exact sciences, the definitely existing objects of research (e.g., the ball lightning), with traceable but inexplicable external indicators of processes prevailing in them, are referred to as the ‘black boxes’. The global economy, obviously, fully corresponds to the definition of the ‘black box’. At the same time, the ‘white boxes’ in the exact sciences are understood as different objects of research with well-known principles of operation, which are the mathematical counterparts of the ‘black boxes’.

As shown below, the ‘white box’ for the ‘black box’ of economic theory is the electric circuit theory where processes are described with differential equations. As for the ‘black box’ of the basic economic process ‘goods-money-goods’, the ‘white box’ is the process in the electric oscillation circuit (Fig. 1b) which consists of an induction coil L and a capacitor C [7], [8].

Indeed, processes in the induction coil L are described with the formula

dt

)t(dIL)t(U L

L (1a)

or an inverse formula

t

0LL dt)t(U

L

1)t(I (1b)

where LU is voltage drop at the inductance coil; LI is the current in the inductance coil;

L is the inductance value; t is time.

At the same time, the behaviour of the vendor in the market is described with the formula

dt

)]t(P)t(Q[dT)t(M VV

VV (2a)


t

0V

VVV dt)t(M

T

1)t(P)t(Q (2b)

where VM is the amount of money (or other means of payment) the vendor received for the goods sold;



VQ is the quantity of the goods sold by the vendor; VP is the price of the goods sold by the vendor;

VT is the production time per unit of the goods sold by the vendor; t is time.

Therefore, the price of the goods set by the vendor equals to the cost of production divided by the quantity of manufactured goods.

Processes in the capacitor C are described with the formula

t

0CC dt)t(I

C1

)t(U (3a)


dt

)t(dUC)t(I C

C (3b)

where CU is voltage drop at the capacitor; CI is the current in the capacitor; C is the value of the capacitor.

At the same time, the behaviour of the buyer in the market is described with the formula

t

0BB

BB dt)t(P)t(Q

T

1M (4a)


dt

)t(dMT)t(P)t(Q B

BBB (4b)

where BM is the amount of payment means (money) the buyer spent on the purchase; BQ is the quantity of goods purchased by the buyer;

BP is the price of the goods purchased by the buyer; BT is the service time per unit purchased by the buyer.

Therefore, the expenses of the buyer equal to the quantity of the purchased goods multiplied by the price per unit.

As can be seen, a perfect mathematical analogy is observed.



www.elvedit.com 176

FIGURE 1. SIMPLEST OSCILLATION LINKS IN ECONOMICS AND RADIO ELECTRONICS [9]

Therefore, for the electric oscillation circuit (Fig. 1b) which consists of a series inductance coil L and a capacitor C , based on the second Kirchhoff’s law, we get the expression

0dt)t(IC1

dt)t(dI

Lt

0C

L (5a)

Differentiating it, we get the differential equation describing processes in the electric circuit under consideration

0LC

)t(I

dt

)t(Id C2

L2

(5b)

For a similar economic circuit, which includes a series (see Fig. 1a) vendor and a buyer, based on the economic interpretation of the second Kirchhoff’s law, we get the expression

0dt)t(P)t(QT

1

dt

)t(P)t(dQT

t

0BB

B

VVV (6a)

Differentiating it, we get the differential equation describing processes in the economic circuit under consideration

0TT

)t(P)t(Q

dt

)]t(P)t(Q[d

VB

BB2

VV2

(6b)

Taking into account that the electric current flowing through the series electrical elements and (Fig. 1b) is the same, i.e. )t(I)t(I)t(I CL , the differential equation (5b) can be simplified to

0LC

)t(I

dt

)t(Id2

2 (7a)

Similarly, in view of the same commodity-money flow though the elements of the economic circuit (Fig. 1a) )t(P)t(Q)t(P)t(Q)t(P)t(Q BBVV , the differential equation (6b) can also be



simplified to

0TT

)t(P)t(Q

dt

)]t(P)t(Q[d

VB2

2 (7b)

The solution of the linear differential equation (7b) is sinusoidal oscillations. Therefore, the ‘goods-money-goods’ process is an oscillation one, which was to be expected, because in the isolated economic circuit under consideration the vendor and the buyer periodically exchange goods for money, and vice versa.

2. Mathematical analysis of structural defects of the capitalist economy: In fact, the process discussed above is only potentially an oscillation one. It is not known in economics, because it has never been implemented yet. Moreover, it cannot be implemented in a random way, just as a house cannot be built at random and vintage vine cannot be made randomly. We must know how to do it.

The non-linear factor. Unfortunately, it is not easy to use the analogy of the electric and economic circuits and processes. This is why we must be very careful when it comes to the practical implementation of the analogy in order not to upset it by any incorrect actions [9], [10].

FIGURE 2. AN EXAMPLE OF CONTENT OF THE ECONOMIC OSCILLATION PROCESS [10]

In the electric oscillation circuit (Fig. 1b) the oscillation process, at any of its phases, has the same physical meaning determined by the motion of electrons. As for the economic oscillation link (Fig. 1a), each oscillation period includes similar processes, which, however, differ by content and follow each other in a certain order. For example, first (Fig. 2a), the vendor may be an



www.elvedit.com 178

employer and the manufacturer of the goods, and the buyer may be an employee who produces the goods. Then (Fig. 2b) the employer pays the employee their wages for the work done. After that (Fig. 2c), the employee turns into a buyer and pays the vendor the price of the purchase. Finally (Fig. 2d), the vendor supplies the purchased goods to the buyer.

Many other variants of organizing the relationship of the buyer and the vendor are possible, especially given that the actual economic oscillation links are not isolated, and, therefore, in the corresponding multi-link and multi-related economic system various economic operations may be performed in different economic links.

However, this is not enough. Even if the procedure discussed above is observed, the ‘goods-money-goods’ process in the economic link can be an oscillation one, but not a sinusoidal one, if all payments by the buyer and good supplies by the vendor are not forced (for instance, on a daily basis, managed by a computer) according to the sine law. Otherwise, the process would be non-linear.

As for processes currently prevailing in the capitalist economy, they are definitely non-linear, because they do not meet the requirements discussed above.

FIGURE 3. NON-LINEAR ISOLATED LINKS [10]

This is true for the economic oscillation circuit under consideration (Fig. 1a), which, in fact, has the form shown in Fig. 3a. Fig. 3b demonstrates its radio electronic counterpart. However, contrary to Fig. 3b, Fig. 3a does not have a non-linear element (a diode); it has a non-linear factor, which stems from disregard of the requirements for linearization of the economic process discussed above.

With respect to the foregoing, a natural question may arise – why do we need all these complications? Does the economy actually need all these harmonic oscillation processes? It turns out that they are necessary, because only they can be the basis for developing the economic theory belonging to the exact sciences. It is impossible to stabilize a non-linear and extremely complicated economic system, i.e., to protect it from economic crises.



Moreover, sinusoidal oscillation processes provide for significantly more efficient use of payment means than the current economy. Indeed, with the conventional methods of payment in the form of discrete (e.g., monthly, or as necessary) payments, money works inefficiently, because there is always either surplus or shortage of it to solve the current economic tasks. If sinusoidal schedules of payments and deliveries were used in economic oscillation systems, means of payment would be working continuously and entirely, and, thus, would bring more profit.

Adam Smith’s ‘invisible hand’. Unfortunately, the problem of revealing structural defects of the economy is not confined to the non-linearity of the actual economic potentially oscillation link. There is one more problem. It is referred to as the ‘invisible hand’ of Adam Smith [11].

The matter is that, in accordance with the Cobb-Douglas production function [12] KALQ , the production volume depends on two slowly varying production factors: (labour costs) and (capital costs). Since this is a steadily increasing function, crises are not supposed to occur in the economy at all.

Therefore, it is obvious that economic processes are influenced by the third factor, more powerful than the two accounted for in the Cobb-Douglas function. Economic scholars often refer to this third factor as the ‘invisible hand’, using the term introduced by Adam Smith [13]. However, all their attempts to identify it have failed.

Then, it is logical to assume that if this factor has not been detected among the objective reasons, it must be subjective, i.e., it is the collective human factor. This is why it manifests itself in the capitalist democratic society with its numerous civil freedoms, and, on the contrary, it was not observed in the socialist totalitarian society where these freedoms were suppressed. The same conclusion was made by Kenneth Arrow [14] and Allan Gibbard [15].

FIGURE 4. FUNCTIONAL SCHEME OF THE SIMPLEST ECONOMIC OSCILLATION LINK ADJUSTED FOR THE HUMAN FACTOR [9], [10], [11], [17], [18]

Indeed, the economic oscillation circuit plotted in Fig. 1a actually includes not only the vendor and the buyer, but also the corresponding human factors they introduce (see Fig. 4), because the actual market participants are ordinary people with common human foibles, habits and other peculiarities. Therefore, they are not always reliable, sometimes they are forgetful, often prone to emotions, illnesses, other random factors and unforeseen circumstances.



www.elvedit.com 180

In view of the aforementioned circumstances, the process in the actual economic oscillation circuit will be described not with a linear differential equation with constant coefficients (7b), but with a linear differential equation with random coefficients (or, in other words, with a parametric differential equation)

0TT

)t(P)t(Q)t(H

dt

)]t(P)t(Q[d)t(H

VBB2

2

V (8)

where )t(HV is the human factor taking into account the behaviour of the vendor,

)t(H B is the human factor taking into account the behaviour of the buyer. At that, )t(HV and )t(H B are random functions of time. Therefore, the solution of the

differential equation (11) is also a random function of time. Since the global economic system is described with a system of such (in fact, even more complicated, because more economic factors must be taken into account) parametric differential equations, it is impossible to predict its development. Consequently, crises in the contemporary capitalist economy are inevitable.

III. CRISIS-PROOF MANAGEMENT OF THE CAPITALIST ECONOMY In order to combat crises in the capitalist economy, the influence of the human factors must,

obviously, be minimized, i.e., it is necessary to provide for const)t(Hlim V and const)t(Hlim B .

By the way, fulfilment of the conditions const)t(Hlim V and const)t(Hlim B allowed eliminating economic crises in the socialist countries. This fact can be considered an experimental evidence justifying the conclusion.

The human factors and may be both internal and external.

The internal human factors are understood as spontaneous unpredictability of behaviour of market participants due to their unreliability, illnesses, forgetfulness, imperfections of contracts or verbal arrangements regulating their activities, rumours, panic, and other similar reasons.

The external human factors are understood as the unpredictable behaviour of market participants determined by the external influence of other individuals or corporate bodies – competitors, public officials, criminal structures and other similar reasons.

For instance, Isaac Newton wrote that simulating people’ s behaviour is a much more complicated task than predicting planetary motion [16].

Therefore, it is obvious that, in, order to minimize the influence of the human factors in the market capitalist economy, some new economic tools are necessary, because the existing economic tools have not been able to offset it. These tools must be different for the internal and



the external human factors.

A. Business-interfaces

We shall refer to a business-interface [19] as a new economic tool intended to eliminate the internal human factors. Contrary to existing contracts, business-interfaces not only regulate the relations of business-partners in detail, but also provide for some new mutual obligations.

The term is borrowed from computer engineering, where an interface is understood as a hardware and software means of connecting various functional elements forming arbitrary complex devices.

Thus, we shall understand a business-interface as a commodity-money means of connecting business partners, which is regulated by the corresponding documents to the slightest details and includes, along with the payment and delivery dates, payment amounts, product ranges, penalties and other conventional terms, other means providing for:

•the most possible linearization of economic processes that must be described with linear differential equations with constant coefficients or linear differential equations with variable coefficients (in this case, on condition const)t(Hlim V and const)t(Hlim B ), to which end non-linear elements and factors must be excluded from the economic system;

•temporal variation of the cash flow and good deliveries as close to the sine law as possible (it can be implemented using the corresponding software).

Obviously, banks must play the primary part in implementation of business-interfaces and, thus, in developing new crisis-proof economy; the activities of banks will be changing the economic outlook of their clients, as well.

Two particular examples of business-interfaces are suggested in [8], [9]. However, there can be at least several hundred business-interfaces corresponding to various economic situations.

Implementation of business-interfaces in economics may bring up the question of whether their use can lead to the excessive regulation of economic activity, in particular, to the suppression of rights and freedoms (similar to socialism). The question is quite natural. The answer is – no, it cannot, because business-interfaces will operate:

•only for the term of a transaction, i.e., from the moment it is made to the moment it is settled;

•only to the extent of the transaction; •only for the business partners indicated in the transaction.

The latter, indeed, will have no freedom of disregarding the terms of the transaction; they will be committed to settle it, which will make the economy predictable and, eventually, facilitate the development of anti-crisis trends. In this respect, it is possible to use the following comparison: nature, providing for the variety of life forms, left all living beings no choice of disregarding their



www.elvedit.com 182

obligations of breeding, and solved this task quite ingeniously. People must just as well demonstrate their ingenuity in developing the business-interfaces they need.

Thus, the economy reformed as suggested above will become both capitalist and socialist: at work, people will abide by the ‘socialist’ discipline, whereas outside work they will be completely free in a ‘capitalist way’.

B. New global computer network

In order to minimize the influence of the external human factors, another economic tool is suggested, namely, the new global computer network TV•net [20] – [23], which is free from all the shortcomings of the Internet. This computer network will enable businesspeople to establish business connections without having to resort to third parties. Due to this, businesspeople will have guaranteed confidentiality of their business connections, and will be able to avoid the unwanted influence of any third parties, individuals or corporate bodies, upon their business.

Unfortunately, the global computer network Internet is hardly suitable for this purpose due to its numerous serious shortcomings. Indeed:

•the Internet does not provide for guaranteed information security, i.e., protection from computer viruses, cyber espionage, hackers and other network threats;

•the www service of the Internet does not allow obtaining and using any valuable information necessary for business and other intellectual activities, and, on the contrary, dumps a lot of junk information upon its users;

•it takes a lot of time to retrieve in the Internet any serious information (if it is there at all) necessary for business and intellectual activities;

•copyright and proprietary rights are often likely to be infringed; •the Internet makes its users constantly purchase new short-lived software.

As for the global computer network TV•net, on the contrary, it will be free from all the shortcomings of the Internet:

•it will provide its users with complete and guaranteed information security due to absence of packet-switched communication and use of one-way broadband communication lines, as well as full control of its owners over the information uploaded in the data bases;

• it will provide its users with the most complete and quickly retrieved information by subscription (it will not be searched for in the Internet);

•it will provide for the utmost confidentiality of the users’ queries, because all information is broadcast simultaneously to all users (and they get it from the broadcast automatically using the corresponding selector software).

1. Business-oriented services: Business-oriented services of the TV•net computer network can be implemented quite quickly and at low cost, because their deployment requires mostly institutional activities. Almost everything necessary for it, apart from a small number of additional applications and some simple devices, is already available at the market. TV•net can



be implemented as a regional network anywhere, where there are television broadcasting networks, because instead of packet-switched communication it uses single one-way communication lines (television and/or fibre optic). High fidelity of information transmission is provided for by anti-noise coding, which has proven its efficiency in deep space communication.

FIGURE 5. SIMPLEST IMPLEMENTATION OF THE TV•NET INFORMATION NETWORK WITHOUT FEEDBACK COMMUNICATION [10], [22]

The trading service broadcasts to the buyers subscribers of the TV•net information on the advertised goods and services, specifying all the details important for the buyer. This initial information is uploaded into the database via any communication lines (including the Internet) by the vendors, who are also subscribers of the TV•net. Advertisements on any other types of business partnership can also be submitted to the database.

In the database, the obtained information is processed and transmitted via a TV adapter, where it is properly encoded, to the TV transmitter of the respective region. The TV transmitter broadcasts this information regularly to all the TV•net subscribers of the corresponding region (this connection is not shown in Fig. 5, and further in Fig. 6, 7 for simplicity). Besides, the same information is fed to users of other regions via satellite repeaters (ground transmitter, satellite, ground receiver).

Users receive the information broadcast via the TV•net to their PCs through TV adapters, similar in terms of their function to modems in the Internet. Subscription to the TV•net is completely identical with that to the pay-to-see television. The only difference is that instead of the TV channels the users indicate the headings they are interested in, based on the respective classification system. The TV adapter extracts the information under the corresponding headings from the received signal and feeds it to the user’s PC, where it is stored in the personal memory and can be browsed through whenever it is convenient for the user. Having reviewed the information, the user can make their choices based on the relevant criteria, either manually or



www.elvedit.com 184

using certain applications, and makes their purchases.

FIGURE 6. SIMPLEST IMPLEMENTATION OF THE TV•NET INFORMATION NETWORK USING FEEDBACK LINES

At the same time, it is noteworthy that users are connected to the TV•net via a single one-way communication line, and are not connected to the Internet. This guarantees absolute information security. Similarly, watching TV bears no threats. Therefore, hereinafter these users will be referred to as the protected users, as opposed to the unprotected users connected to the Internet. In case a protected user needs to submit any information via feedback lines through the Internet (e.g., order a door-to-door delivery), they must use another (i.e., the unprotected) PC (see Fig. 6). However, the protected PC and the unprotected PC must not be connected to each other with any communication line (at least, not with a bidirectional one).

The exchange service operates in a way similar to the trading service, with the only difference: it is not fixed market, but auction price market. Therefore, it is necessary to use the variant of the TV•net implementation with feedback lines (Fig. 6).

The administrative service of the TV • net computer network provides for guaranteed information security of banks, state institutions, corporations and any participants, who are now quite vulnerable to the network threats of the Internet.

In large institutions, many protected computers receive information; therefore, it is expedient to connect them to the TV•net information network via a local area network (Fig. 7). At the same time, the protected PCs connected to the TV•net and the unprotected PCs connected to the Internet must not have inter-computer communication, although they may be used by the same users and be located on the same desktop.

2. Intellectually oriented services: The objective of human intellectual activity is to reveal, based on the available information, the trends (e.g., in administrative or economic management)



and regularities (e.g., in science) in the situations and processes under investigation. These are the tasks solved by the analytical services of banks, corporations, intelligence offices, military staffs, as well as by scientists.

FIGURE 7. SIMPLEST IMPLEMENTATION OF THE TV•NET INFORMATION NETWORK

USING LAN AND FEEDBACK LINES [10]

Correct assessment of situations and relevant managerial decisions depend on the results of analytical activities. Analytical service is just as important in scientific work. Since human rational thinking is low-factor, so far almost all known natural laws, with a rare exception (here belongs, for instance, Newton’s version of Kepler’s third law) are described with functions of not more than three variables. However, nature cannot be so primitive as to have only these simple regularities. Actually, it is the people who are unable to comprehend more complicated regularities with their intellect. You can make sure this is really so – just try to imagine any multi-dimensional object, for example, a four-dimensional cube (a tesseract or octachoron).

These tasks are solved by the analytical service discussed below, as well as the educational service, because developmental teaching is an integral part of fostering creative thinking.

The analytical service is intended to help users in revealing trends and regularities. Therefore, working in a way similar to the trading service (Fig. 5), the analytical service also supplies its subscribers with information, but of a different kind. However, being within the framework of the usual rational low-factor thinking, there is not much it can do to solve the problem, except for interpolation and extrapolation algorithms, algorithms for revealing correlation relationships and some others. Although, it will be quite useful in this respect, anyway.

However – and this is most interesting – the analytical service will allow implementing the multi-factor human-computer super-intelligence, which will make it possible to reveal multi-



www.elvedit.com 186

factor trends and regularities. Certainly, this will enable users to make decisions that are more efficient in their sphere of activity.

Let us specify that human-computer super-intelligence [24] – [26] is understood as a new task alternative to the deadlocked task of creating artificial intelligence that aimed, in particular, at developing a sort of computer super-intelligence based on the computer thinking emulating human rational thinking.

Thus, let us give our reasoning. At present, a conventional definition of intelligence – either human or computer – still does not exist. Therefore, we will understand human intelligence as a set of subsystems, which includes low-factor rational thinking (active in the waking state), multi-factor unconscious thinking (active round the clock, but most of all in the state of sleep), and a number of other subsystems. The major part in this set is played by the subsystem of multi-factor subconscious thinking. It is referred to as subconscious for the reason that many people do not even realize it exists. Many other people do not understand what it is for. People do not even know why they (and not only they, but also all living beings) sleep, because they believe that in sleep they are uselessly idle.

However, nature does nothing in vain. This is why nature makes people sleep, in order to make it possible for the multi-factor unconscious thinking to process the information accumulated in the waking state by turning off all information input and output senses, as well as the low-factor rational thinking (so that they do not interfere). This is why there is a saying: “If you have a problem, sleep on it”. This is why, when someone is sick, his or her temperature is lower in the morning than at night. This is why many scientific discoveries were made in sleep. Thus, if people did not sleep, they would not survive as a biological species.

As for the much more primitive, but much faster low-factor rational thinking, nature gave it to people to use in active everyday routine for life support – getting food, performing some work, bringing up children, defending themselves from enemies, and so on. It is referred to as low-factor because it is based on processing of visual images, which are not more than three-dimensional. However, almost all real-life processes (in economics, medicine, operation of scientific and technical systems, making weather forecasts and so on) depend on a large number of factors.

At the same time, artificial intelligence tries to emulate human low-factor rational thinking. Nevertheless, eventually, it was challenged with an ambitious task of excelling human intelligence, i.e., computers were to be taught to solve intellectual tasks without a human, instead of a human and better than a human. However, this is a utopia. Actually, over the decades of research devoted to artificial intelligence, scientists have not even managed to teach a computer to tell a dog from a cat.

This is why, in view of the phenomenal successes of engineers and scientists in the advancement of computers, it is time to formulate a new problem – the problem of development of human super-intelligence, which is understood herein as the development of human-computer systems capable of solving multi-factor tasks.

The matter is that the human intelligence created by nature was not intended to solve



scientific and other intellectually demanding problems – thousands years ago people were busy with other things. However, if people turned out to be able to solve intellectually demanding tasks, this testifies only to the fact that human intelligence is highly advanced and has huge (basically unexplored) possibilities for further development, both by means of intellectual training (which is constantly done by scientists) and by means of additional use of computer resources (which is suggested herein).

Thus, in terms of implementing human super-intelligence, we are interested in the possibility of intentional, contrary to the process of unpredictable intellectual insight, use of multi-factor unconscious human thinking in the waking state. How can a computer be useful in solving the problem? In order to answer the question, let us analyse the way a scientist thinks. It is easy to see that they perform a repeated cycle of the following operations:

•first, they gather all the information available, study it and single out the significant factors relevant to the problem under investigation;

•then they look for a mathematical dependence among the significant factors revealed before and the desired result;

•then they test the revealed dependence by using it to explain the known experimental data and to perform new experiments, which are supposed to confirm the mathematical dependence under investigation;

•if the results of the new experiments do not agree with the mathematical dependence, further research of the available data is performed, and the significant factors are reviewed;

•and so on.

As can be seen, the stages of information gathering (both in a library and in the Internet) and detection of significant factors are quite tedious, even more so that the necessary information may turn out to be in the field of other sciences. At this stage, the analytical service of the TV•

net computer network can be quite useful.

At the next stage of searching for mathematical dependence, a computer may be even more useful, because a human cannot comprehend functions of more than three variables.

Finally, at the last stage of harmonization of the revealed mathematical dependence with the experimental data, a computer is also very useful, because it can quickly review all possible variants and draw the attention of the researcher to the circumstances that need to be given special consideration.

However, at any of the stages a computer cannot operate without a human, because only people are able to solve indeterminate problems (let us recollect the task of telling a dog from a cat).

The educational service is also an intellectually oriented one [27], first, because creative thinking is fostered in the process of training, and, secondly, because creative people often have to get additional training or re-training in the course of their professional life.

Thus, taking into account that intellectual work is getting more and more demanded in the modern society, the primary objective of education must be the intellectual development of



www.elvedit.com 188

people.

The educational service that solves the problem operates as follows. Using the computer-television broadcasting network – the TV•net (see Fig. 5), its users, independently or via their educational institutions, receive and download to their PCs:

•textbooks and work-books supplied with a large number of hyperlinks to other sections of textbooks (other textbooks, as well) and FAQ sections;

•supplementary further reading; •problem books with detailed solutions of typical problems and advanced problems; •learning and development and other software.

Having received the information, the users master the material on their own or with the help of a tutor. At that, if users have questions that cannot be answered with the help of either a tutor or the FAQ section, they can submit their questions to the database via feedback lines (Fig. 6) and get the necessary explanations.

IV. SUMMARY Thereby, it is possible to develop economic theory that meets the criteria of the exact sciences.

It is also possible to create the global crisis-proof economy.

However, economic scholars are not able to do it on their own, without the help of radio electronics, information technologies, and software development professionals. This problem must be solved by consolidated efforts.

REFERENCES

[1] Bannock G. and Baxter R. (2009). The Palgrave Encyclopedia of World Economic History: Since 1750. Palgrave Macmillan, Basingstoke.

[2] Antonov A.A. (2013). Unpredictable discoveries. Lambert Academic Publishing. Saarbrücken.

[3] Neumann J. von and Morgenshtern O. (1947). The Theory of Games and Economic Behavior. (2nd ed.) Princeton University Press. Princeton.

[4] Bruno M. and Esterly W. (November 1994). Inflation Crises and Long-Run Growth. World Bank.

[5] Polterovych V.M. 2007. Elements of the reforms theory. Ekonomika Publishing. Moscow.

[6] Antonov А. А. (2013). Obsolete scientific dogmata hamper development of human civilization. European Journal of Academic Research. 1(1), 22-30.

[7] Antonov A.A. (2010). Differential equation for the ‘goods-money-goods’ process. European Journal of Scientific Research. 40(1), 27-42.

[8] Antonov A.A. (2010). Economic oscillating systems. American Journal of Scientific and Industrial Research. 1(2), 359-363.



[9] Antonov А.А. (2013). Сoncept of crisis-proof economy. European Journal of Academic Research. 1(1), 31 -38.

[10]Antonov А.А. (2013). Fundamentals of crisis-proof economics. International Journal of Innovation and Applied Studies. 2(3), 196-215.

[11]Antonov A.A. (2013). Discovering of Adam Smith’s ‘Invisible Hand’. International Journal of Managment, IT and Engineering, 3(1), 1-10.

[12]Filipe J. and Adams G. (2005). The Estimation of the Cobb Douglas Function. Eastern Economic Journal, 31(3), 427-445.

[13]Smith Adam. (1776). An Inquiry into the Nature and the Causes of the Wealth of Nations. In Cannan E. (Ed.) 1977. University Of Chicago Press.

[14]Arrow K. J. (1963). Social Choice and Individual Values. (2nd ed.) Wiley: NY.

[15]Gibbard A. (1973). Manipulation of voting schemes: a general result. Econometrica, 41(4), 587-601.

[16]Bouchaud J.P. (30 October 2008). Economics needs a scientific revolution. Nature. 455, 1181

[17]Antonov А.А. (2012). New anti-crisis instruments for market economy. ARPN Journal of Science and Technology, 2(8), 738-744.

[18]Antonov A.A. (2011). Realization of Crisis-Free Economy. International Journal of Emerging Sciences, Special Issue: Selected Best Papers, 1(3), 387-399.

[19]Antonov А.А. (2009). Safe Global/Regional Informational Network. European Journal of Scientific Research, 28(1), 165-174.

[20]Antonov А.А. (2012). The new global information network free from the drawbacks of the Internet. ARPN Journal of Science and Technology, 2(10), 957-962.

[21]Antonov А. А. (2013). New Business-Oriented Global/Regional Information Network. International Journal of Business Information Systems, 12(3), 321-334.

[22]Antonov А.А. (2013). New Global Сomputer Network. International Journal of Managment, IT and Engineering, 3(1), 11-22.

[23]Antonov A. A. (2010), Human-computer super intelligence, American Journal of Scientific and Industrial Research, 1(2), 96-104.

[24]Antonov A.A. (2011). Human Super Intelligence. International Journal of Emerging Sciences, 1(2), 164-173.

[25]Antonov A.A. (2011). From artificial intelligence to human super-intelligence. International Journal of Computer Information Systems, 2(6), 1-6.

[26]Antonov А.А. (2011). Realisation of Human Super-Intelligence: Developmental Learning. WSEAS Transactions on Advances in Engineering Education, 8(4), 109-119.



www.elvedit.com 190


Alexander A. Antonov received PhD degree in Radio Electronics at Saint-Petersburg State University of Aerospace Instrumentation in Russia. He was Associate Professor of Tula State University in Russia and Leading Scientific Officer of Information Recording Institute of the Ukrainian Academy of Sciences. Now he is Director of Research Centre of Information Technologies “TELAN Electronics” in Ukraine, full member of Russian Physical Society, member of International Optical Society SPIE, author almost 200 patents, several dozens of scientific papers and 3 books.


E-Business Qualitative Criteria ApplicationModel: Perspectives of Practical

Implementation

Authors





Abstract

Constantly changing business environment makes the traditional business switch to electronic. One of the main problems in the development and implementation of e-business is e-business qualitative criteria uncertainty. Quality is a very important objective for both – business and customers. But there are no e-business qualitative criteria centrally and systematically analysed and defined in the theory, the selection as well as the evaluation of these criteria are not clear. There is discussed and analyzed question of creating e-business qualitative criteria In this paper. The aim of the paper is to create e-business qualitative criteria, to analyze the possibilities of their application and propose e-business qualitative criteria application model. The objectives are – to analyze theoretical aspects of e-business qualitative criteria creation and application; carry out a qualitative survey of e-business experts and analyze it’s data; analyze e-business qualitative criteria application model implementation possibilities and perspectives. Theoreticalaspects of e-business qualitative criteria include e-business qualitative criteria formation guidelines. Therewere defined 4 e-business qualitative criteria: matching the value curve; orientation to the customer;information and data quality; creativity. The paper relies on scientific literature analysis, the qualitativeresearch method and the method of dynamic modeling are applied as well. Also, there is carried out thetheoretical narrative, systematic, comparative analysis. After analyzing theoretical aspects of e-businessqualitative criteria, conducting e-business experts qualitative opinion survey and proposing e-businessqualitative criteria, there was created a model of their application. E-business qualitative criteria applicationmodel includes the input – start of e-business, answering to added value curve questions, customersatisfaction analysis, the information, data quality analysis, level of creativity. After analyzing business inaccordance with all e-business qualitative criteria, it can be seen what needs to be improved, and thedirection in which to do so, because the output (high-quality e-business) will be achieved only when e-business in great extent or completely satisfy these criteria.

Key Words

E-business, E-business qualitative criteria, E-busines qualitative criteria creation, E-busines qualitativecriteria application model.



192

I. INTRODUCTION Retail market will change more in the next five years than have changed over the last

twenty. These changes lead to full e-business development [6]. Over more than a decade e-business has become a rising phenomenon that has affected many industries structure [10]. Businesses simply must transform to e-business in order to survive in the knowledge-based economy and market [22]. For the companies in the age of the new economy it is important that elements of e-business system would be included in their management system [21]. Constantly changing business environment makes the traditional business switch to electronic. But there are not created e-business qualitative criteria, possibilities of their application are undetermined. In this article there will be discussed a little-analyzed question of e-business qualitative criteria creation and application.

Scientific issue. One of the main problems in the development and implementation of e-business is e-business qualitative criteria uncertainty. Quality is a very important objective for all business and customers. Quality is exclusivity, differentiation of e-business [19]. Without knowing the exact e-business qualitative criteria, it is difficult to develop the business itself, it is difficult to determine, what is needed to achieve and what criteria should be assessed by business. According to this, it is possible to determine the derivative problem –in academic sources and practice there are not generally set e-business qualitative criteria, there is no analysis of their application.

Object of the research. E-business qualitative criteria creation and application.

Purpose – after analyzing theoretical aspects of e-business qualitative criteria creation and application, to create e-business qualitative criteria, analyze the possibilities of their application and propose e-business qualitative criteria application model.

There have been set the following objectives for the above-mentioned purpose to be achieved: 1. To analyze theoretical aspects of e-business qualitative criteria creation and

application; 2. To carry out a qualitative survey of experts and analyze it’s data, which, through

expert knowledge, will let find out the main issues arising from the implementation of e-business, choose the criteria that can be used for measuring e-business quality;

3. To propose e-business qualitative criteria application model.

Methodology – the paper relies on scientific literature analysis, the qualitative research method and the method of dynamic modeling are applied as well. The work carried out theoretical narrative, systematic, comparative analysis. A qualitative expert opinion survey was carried out interviewing 9 e-business experts. The experts’ opinion was sought to analyze by a standardized interview or questionnaire form. Kendall’s concordance coefficient was calculated by Statistical Package for the Social Sciences (SPSS) program. After analysis of e-business qualitative criteria creation and application, there was created a model.

Practical significance. Practical significance reflects analyzed and experts distinguished the most important e-business qualitative criteria, suggested model of e-business qualitative criteria application. This information can be used in the implementation of quality-oriented e-

E- business Qualitative Criteria Application Model: Perspectives of Practical Implementation Tadas Limba, Gintarė Gulevičiūtė


193

business. E-business quality, qualitative criteria are little researched area, so the results can be the basis for more efficient e-business development.

II. E-BUSINESS QUALITATIVE CRITERIA FORMATION GUIDELINES In terms of e-business quality (focusing on the purchase of goods and services in electronic

shops) there could be identified two important questions [23]:

What are the dimensions of quality, features that attract users to the website and make them come back again?

What are the steps to carry out for the business to ensure, that their websites would differ from their competitors, so that users could be sure that they are getting greater value from them?

It is important to focus attention on the customer's perspective, highlight the key requirements to ensure customer satisfaction [23]. These requirements is defined in figure below (see Figure 1). It is argued that customer satisfaction is based on the following aspects:

Easy to use (Website Design) How does the website look like? Costumer confidence (How is guaranteed?) Direct resources (ability to offer and deliver products and services) Bridging services (how is interacting with customers and maintaining their loyalty?)

FIGURE 1. CUSTOMER SATISFACTION: QUALITY DIMENSIONS. SOURCE: MOHANTY ET AL., 2007, P. 224-237



194

There are three types of requirements that must be met by e-business in order to fit the needs of customers. There are performance-expected, basic-must requirements that the customer immediately expect these requirements to be fulfilled by e-business and delight-excitement features, that creates an additional, more satisfaction to the customer. Currently, customer orientation is the basis for achieving business goals. It is believed, that interacting with customers helps to expand the profitable business opportunities [5].

In the rapidly growing e-business, organizations in the open environment are required to cooperate to achieve certain goals related to their business model. In such an open environment, the privacy of organizations becomes a critical challenge [1]. In terms of e-business quality, the quality of information is important, because it determines customer perception of goods or services quality. There are presented the data quality dimensions [18], (see Table 1).

TABLE 1. DATA QUALITY DIMENSIONS. SOURCE: KORONIOS, XU, 2005, P.74.

Dimension Definition Accessibility The extent to which data is available, or easily and quickly

retrievable. Appropriate amount of data

The extent to which data is appropriate for the task.

Believability The extent to which data is regarded as true and credible. Completeness The extent to which data is not missing and is of sufficient breadth

and depth for the task. Concise Representation The extent to which data is compactly represented. Consistent Representation

The extent to which data is presented in the same format.

Ease of Manipulation The extent to which data is easy to manipulate and apply to different tasks.

Free-of-Error The extent to which data is correct and reliable. Interpretability The extent to which data is correct and reliable. Objectivity The extent to which data is unbiased, unprejudiced, and impartial. Relevancy The extent to which data is applicable and helpful for the task at

hand. Reputation The extent to which data is highly regarded in terms of its source or

content. Security The extent to which ascess to data is restricted appropriately to

maintain its security. Timelinesss The extent to which data is sufficiently up-to-date for the task at

hand. Understandability The extent to which data is easily comprehended. Value-added The extent to which data is beneficial and provides advantages from

its use.

These dimensions of quality are the main requirements to be met by the data, information in e-business – to be reliable, quick to find, easy to understand, helping to carry out the task. There can also be identified 3 dimensions of information quality [15] (see Table 2).



195

TABLE 2. INFORMATION QUALITY DIMENSIONS. SOURCE: JIN KIM ET AL., 2005, P. 78.

Dimension Meaning Constructs Content (Product) Aspect of Information Quality

Content This dimension deals with the intrinsic information content issues that are geared toward providing users with accurate, relevant, and complete information, thereby addressing primarily problem of irrelevant information in e-business systems.

Information Accuracy: Freedom from mistakes in the information content and hyperlinks provided within Web pages. Information Relevance: Pertinence to users' interests of the information content and hyperlinks provided within Web pages. Information Completeness: Availability as needed of the information content and hyperlinks within Web pages for users to complete specific tasks in an effective manner.

Presentation and Delivery (Service) Aspects of Information Quality Form This dimension deals

with information presentation issues that are geared toward enhancing users' cognition, thereby primarily addressing the problem of cognitive overhead.

Interface Structural Quality: Primarily comprises interface consistency and structural awareness. Interface implies consistency in the structural arrangement and style of information content and hyperlinks within an e-business application. Information Packaging Quality: Refers to how effectively a variety of information in various media types is packaged within the Web interface for presentation to users. Includes the amount and cohesiveness of information content and hyperlinks presented within the interface, and semantic relationships among them. Information Accessibility: Refers to the ease and efficiency with which a user can navigate within an e-business application to access and retrieve desired information.

Time This dimension deals with information delivery issues that are geared toward providing users' better control over temporal aspects of their actions thereby providing them with a sense of temporal orientation and, thus, addressing primarily the problem of disorientation in e-business systems.

History Maintenance Quality: Refers to the flexibility and comprehensiveness of features that an e-business application provides to its users for specifying and maintaining history of user actions and data and system states of the application. Information Delivery Quality: Refers to the flexibility and comprehensiveness of features that an e-business application provides to its users for specifying and controlling the temporal relationships among the various hypermedia components for effective delivery of integrated hypermedia information to users. Information Currency: Refers to the temporal accuracy of information content and links on Web pages.



196

Information quality dimensions is defined by the content, form, time, however, the composition of information quality dimensions reminds data quality dimensions.

The Internet has brought the free access to information. Information is obtained much faster, but there are problems related to information security and data quality maintenance [8]. In terms of e-business quality stressing the importance of quality, there could be added the system and the quality of services as an important element [28]:

The quality of information - website content, the completeness, clarity, format. The user can get the right information about the product supplier.

System quality - defines e-business systems desired characteristics: usability, reliability, feedback, availability, timeliness.

Service quality - fast responsiveness, reliability, empathy, focusing on consumer in sales of goods or services.

There could be emphasized not only the information and data quality importance in e-

business, but also the importance of safety and security [13]:

Information security - a key tool to remain competitive against the other is to ensure the security, integrity and secure business communications and customer information. The basic principle is to ensure, that any sent information would reach its recipient. Information security is also important in order to gain a competitive advantage in an ethical and legal compliance. Unauthorized communication of content, malicious communication on behalf of organization is only some of the risks related to information security and can cause damage to business reputation, financial loss, loss of confidence and loss of information.

Data protection – e-business must comply with all laws relating to data protection. It should ensure, that the data processing operations would be carried out in accordance with the law : - Fair and lawful processing of data; - Data collection and further processing only to the legitimate objectives of the case; - To maintain personal data collection and processing of the adequacy and

appropriateness; - To maintain accuracy; - To store data for longer than is really needed; - Appropriate measures are taken to ensure data protection; - The data shall not be transferred to third parties, unless they ensure data

security.

Many authors identify the data and information quality importance in terms of e-business quality. This may be one of the most important criteria for analyzing e-business quality.

Another very important aspect, which can be associated with e-business quality - creativity. In a broad sense, creativity can be defined as the ability to develop new ideas [14]. In e-business case analysis, there occur 3 possibilities for creativity to reveal [14]:



197

Comparative analysis of domestic industry (through their culture) – there could be supposed, that e-business in Germany sells a variety of goods online and want to improve their performance. In particular, it will examine other similar e-business situation in Germany. A comparative analysis of their culture will let understand the level of competition, the business will find some creative ideas that enhance the competitive edge.

Comparative analysis of domestic industry (within their own culture) - the same e-business in Germany is looking for new creative ideas and go out of their cultural boundaries. Perhaps looking to countries, where e-business development has more experience, where to find the model, options that are available to customize e-business.

Inspiration that comes from other industries - most significantly more creative ideas comes analyzing e-business that specialize in other activities than the example e-business in Germany, which sells a variety of goods online. There can be analyzed a completely different e-business, such as financial services or media. Clearly adapt best practice in this case is more difficult, requires more effort.

All of these options appear in the figure below (see Figure 2).

FIGURE 2. SEARCH FOR NEW IDEAS WITHIN THE DOMESTIC INDUSTRY AND OTHER INDUSTRIES. SOURCE: JELASSI, ENDERS, 2005, P. 43



198

In 1998 W. Ch. Kim and R. Mauborgne had already written about new business

opportunities to raise the quality. They introduced business value curve (see Figure 3), which is based on the answers to four questions [17]:

What factors need to be reduced? What factors should be raised above the standard? What factors, which are considered to be granted, should be canceled? What should be created, that no one has not proposed so far?

FIGURE 3. VALUE CURVE. SOURCE: KIM, MAUBORGNE, 1998, P. 85

Although the value curve has been introduced and adapted to the traditional business, it can

be used to analyze the modern e-business [7]. The main value delivered from using the internet is improved brand and/or product awareness [30]. Electronic business has dramatically changed consumers’ purchasing and buying behaviors [12]. For example, Amazon can be identified as the creator of the new value curve, because when analyzing Amazon online shop, following four questions can be answered [7]. Amazon has always been focused on providing high quality services to customers [16]. It has expanded the range of books. Amazon does not limit its strategic partnerships to suppliers and distributors, but also engages in alliances banks that offer Amazon a credit card which also adds value to customer relations [9]. Also, Amazon created helpful reviews system that offers greater potential value to customers [24]. Much of the company's success depends on the fact that it was able to create a strong name of Amazon company [11].

In terms of e-business systems the quality can be emphasized through the following criteria [26]:



199

Activities - from e-business there is expected to respond quickly to user requests, despite the large number of transactions carried out. The activity will also determine the time at which all requests are served. It can be measured in performance, response time, turnaround time, efficiency.

Availability - system is able to provide continuous service to its customers. The term is related to system failures and how quickly the failure is removed.

Scale - this is the system's ability to expand its operations in accordance with the requests of customers, transactions, while maintaining the same performance. This is the system's ability to adapt to economic growth. It is important that the e-business model could be expanded, because there is a lot of growth potential in the online market.

Security - concerns the confidentiality, authentication, authorization, encryption and access control. E-business can apply different security policies depending on the service requests. There is a growing concern about security in e-business, where the service is delivered over the public internet.

Integrity - data plays an important role in e-business operations. Data integrity has gained considerable importance. Preserve the integrity of data is very important. Honesty is the system's ability to prevent unauthorized access or modification of data. Data integrity can be compromised when the system crashes, errors are committed by individuals, viruses or hackers.

Interoperability - measures the degree to which e-business service can interact with clients and servers and to be implemented in different languages and/or on different platforms.

Fault tolerance - it is the system's ability to operate at partial failure. For operations workloads, it is the system's ability to recover from failure without loss of data or updates of recent transactions.

Competitiveness - it is the system's ability to carry out several independent consumer orders simultaneously. Competitiveness is the capacity of the system to withstand various types of applications of varying load conditions.

Reliability - is the system's ability to keep the load of sequels, working an acceptable level of performance. Reliability depends on various factors, such as memory usage, performance and response time. Reliability is a common tool for the system to maintain quality of the service and is associated with failures per unit of time.

Resistance - measures the degree to which e-business is still able to process properly in the face of misleading input.

E-business is constantly changing and improving, the quality can be assessed with a variety of aspects. There is presented the 10 trends that will affect e-business in the future [3] (see Table 3). Most are specifically targeted to the user's satisfaction.



200

TABLE 3. TOP 10 TRENDS SHAPING THE FUTURE OF E-BUSINESS. SOURCE: BALARAMAN, KOSALRAM, 2012, P. 12-13.

No. Trend What is the Trend? 1. Collective intelligence When a customer visits a site - a shopper's persona is

defined within a few interactions. 2. Social network integration Integrating a "share" button enables users to share

content to their social networking site. 3. Mobile sites A dedicated mobile site experience is a must. 4. Location based tie-ins GPS capabilities of mobile devices usher in a new era

of exciting cross-channel promotion capabilities. 5. Experiential user interface Beyond simply being easy to use, modern ecommerce

sites for innovative brands can be experiential and immersive.

6. Contextual visualization Shoppers increasingly expect to visualize how a product will fit into their life and style.

7. Dynamic grid expansion and liquid layouts

Utilize liquid layouts to automatically size your product display based on the shoppers' resolution.

8. Minimize UI cruft Shoppers come to your site to see your products, not your fancy navigation systems.

9. Rich DHTML and AJAX Instead of having to reload a page every time the shopper clicks, these technologies enable a world of rich interactions (instantaneously).

10. Get textual With the advent of HTML5 and font-serving technologies such as TypeKit, the web designers' typographic palette has been opened up as never before.

Various authors often write about e-business quality. In summary, the most frequently

mentioned criteria relating to e-business is customer orientation; information, data quality; creativity. These criteria in one way or another usually encounter in analyzing the scientific literature related to e-business quality. It is hard to understand, which could be the most important criteria, so in experts survey, they will be asked to pick out, which criteria is the most important.

III. RESEARCH ON E-BUSINESS QUALITATIVE CRITERIA CREATION AND APPLICATION

A. Research Methodology

An expert qualitative opinion survey was carried out in which nine e-business experts were interviewed. There has been chosen qualitative research method, because qualitative research adopters argue that in this way the data obtained further information about the object rather than from quantitative studies [29]. The experts have personally been given a questionnaire, there was a direct interaction about the form-related issues.

The study was conducted by interviewing a variety of e-business experts. There were interviewed 9 following experts:

JSC “Exacaster” Chief Data Analyst; “Adbox” CEO;



201

“Metasite Business Solutions” Data Analyst; “Metasite Business Solutions” Sales Manager; JSC “Pigu.lt” Marketing Project Manager; “Getjar Inc” Advertising Sales Manager; JSC “Antigravity Payment Systems” Marketing and Communications Specialist; JSC “Antigravity Payment Systems” Corporate Manager; JSC “EVP International” Development Manager (for foreign markets).

Selected exactly this number of experts, because it is an acceptable number of experts in the

methodological assumptions formulated in classical test theory. The theory states, that the aggregate decisions reliability and decision-making (in this case, the expert) number links quickly extinguishing a nonlinear relationship (see Figure 4). There is evidence that the aggregate expert assessment modules with equal weights in small group of experts, judgments and assessments do not yield high expert group decision evaluation and accuracy [20, 4].

FIGURE 4. EXPERT EVALUATION STANDARD DEVIATION DEPENDENCE ON THE NUMBER OF EXPERTS. SOURCE: BALEŽENTIS, ŽALIMAITĖ, 2011, P. 25

There could be seen that the accuracy of the estimates and judgments are sufficiently large,

when the number of experts reaches 9. After this number, rising of accuracy is slight so 9 experts is enough to get precise information.

Before analyzing the obtained data there was clarified expert opinions compatibility. Two experts can assess the compatibility of quantitative correlation. If the number of experts are more than two, the group of experts compatibility level indicates Kendall’s concordance coefficient [25]. With Statistical Package for the Social Sciences (referred as SPSS) program Kendall’s concordance coefficient was calculated. If the opinion of experts is coordinated, concordance coefficient W value is close to the 1, if they differ W value is close to 0 [25]. Since resulting figure is closer to 1 than 0, it is concluded that the expert opinion is sufficiently coordinated. The resulting Kendall coefficient of concordance: W = 0,707 (see Table 4).

Number of experts

Standard deviation



202

TABLE 4. TOP 10 KENDALL'S COEFFICIENT OF CONCORDANCE

Sum of Squares df Mean Square Friedman'o

Chi-Square Sig

Between People 4.259 8 .532 Within People

Between Items 46.759a 5 9.352 34.030 .000 Residual 15.074 40 .377 Total 61.833 45 1.374

Total 66.093 53 1.247 Grand Mean = 2.8704 a. Kendall's coefficient of concordance W = .707.

Since the resulting figure is closer to one than zero, it is concluded that the expert's opinions are fairly coordinated.

B. Research Data Analysis

In 1998 W. Ch. Kim and R. Mauborgne has already written of new business opportunities to raise the quality. They introduced business value curve (see Figure 3), which is based on the answers to four questions [17]. The experts were asked whether they agree with the statement that in order to develop e-business and seek for the highest quality, there would be able to use the value curve and create e-business primarily answering to these four questions (see Figure 5).

FIGURE 5. EXPERTS OPINION IF THERE WOULD BE ABLE TO USE THE VALUE CURVE IN ORDER TO DEVELOP E-BUSINESS AND SEEK FOR THE HIGHEST QUALITY

There could be seen that the 5 experts agreed with this statement, 4 partially agreed.

Another relates to the previous question - what else questions should be answered (excluding that 4 in value curve) in order to develop e-business and seek for the highest quality.

54

0

agree partially agree



203

Experts declare that in order to develop e-business and seek for the highest quality, there should be answered and these following questions:

Encouraging. How to make the purchasing process more enjoyable and create natural feeling to the buyer? There have been distinguished a great example of Apple's AppStore or Google PlayStore, where applets acquisition is resolved incredibly easy and comfortable, while the Amazon multisite is still missing it.

Does it solve some kind of problem? Is there a high human cost? Could this be a solution, which will use most people?

How to reduce people's fears? For example, Lithuanians is still afraid to buy online, do not trust e-shops, do not believe that the goods will be delivered on time.

How to make payment convenient? It was emphasized that e-stores that use the One -Click, or Single -Click buying (Amazon, Apple) make the buying more convenient for customers. Also important is the method of procurement and security. Usually customers will choose a store, which allows pay through PayPal.

Educate - how to ensure that the audience understands the value and to eliminate the fear barrier that prevents the audience from buying in e-shop?

Present - how to make everything so attractive, that users cannot resist and purchase product/service?

After considering expert answers, there can be added more questions to the value curve that should be answered in order to develop e-business and seek for the highest quality (see Figure 6).

FIGURE 6. ADDED VALUE CURVE

After analyzing theoretical aspects of e-business quality, there were defined 4 e-business qualitative criteria:



204

Matching the value curve (see Figure 3); Orientation to the customer; Information and data quality; Creativity.

The experts were asked to distinguish, which e-business qualitative criteria is the most

important (see Figure 7).

FIGURE 7. EXPERTS OPINION ABOUT WHICH E-BUSINESS QUALITATIVE CRITERIA IS THE MOST IMPORTANT There could be seen, that most experts (5 experts) believe, that the most important e-

business qualitative criteria is orientation to customer, 3 experts believe that it is also important information and data quality.

The experts survey helped to define, if there would be able to use the value curve in order to develop e-business and seek for the highest quality, what else questions should be answered (excluding that 4 in value curve), which e-business qualitative criteria is the most important.

IV. CREATION OF E-BUSINESS QUALITATIVE CRITERIA APPLICATION MODEL

A. Modeling Methodology

After proposing e-business qualitative criteria, it is useful to create a model of their application. Models can be divided into the mathematical, statistical and qualitative [27] (see Figure 8).

5

3

1

0

1

2

3

4

5

6

orientation to customer

Information and data quality

Creativity



205

FIGURE 8. CLASSIFICATION OF MODELS. SOURCE: SIDEKERSKIENĖ, 2007 In this case, there will be created a qualitative model using the qualitative process

descriptions. In developing any model there is certain rules. Each model's performance is independent, but can work with other processes. Each process consists of a number of resources, activities and information. It is important to determine the pattern of the input and output. No activity will not begin without the inputs, while the output is the result of the model [2] (see Figure 9).

FIGURE 9. THE MODEL STRUCTURE AND INTERACTION. SOURCE: AYTULUN, GUNERI, 2008, P. 2745.

Described rules has been used in creating e-business qualitative criteria application model.



206

B. Model Analysis

E-business qualitative criteria application model (see Figure 10) starts from the input - in this case the entrance is starting e-business. When launching e-business, after forming some idea, firstly it is important to answer to added value curve questions. After answering to these questions there will be clear, how e-business will differ from others, what value will create for the customers and so on. Further there should be carried out customer satisfaction analysis. It is best to do this analysis after creating e-business website, analyzing, how it is focused on the customer. The website should be analyzed by using 24 costumer-oriented criteria. Moreover, the information, data quality analysis can be done. Finally, it is important to determine e-business level of creativity. The highest level of creativity is reached only after a longer e-business existence, but it can be used as a guide to help develop e-business. After analyzing business in accordance with all e-business qualitative criteria, it can be seen what needs to be improved, and the direction in which to do so, because the output (high-quality e-businesses) will be achieved only when e-business in great extent or completely satisfy these criteria.

REDUCEWhat factors need to be

reduced?

CREATEWhat should be

created, that no one has not propose so far ?

RAISEWhat factors should be raised

above the standard ?

CANCELWhat factors which

are considered to be granted, should be

canceled?

ENCOURAGEHow to make the

purchasing process more enjoyable?

SOLVEDoes it solve some kind of

problem?

PAYHow to make payment

convenient ?

EDUCATEHow to ensure that the audience

understands the value and to eliminate the fear barrier ?

PRESENTHow to make everything so attractive, that users cannot resist and purchase product /

service?

Answers to added value loop questions Orientation to customer

Leve

l of c

reati

vity

High

LowE.

business

Comparative analysis of the

domestic industry

Other similar e.

businesses

Selling goods online

Financial services Media Automotive

industryTelecommunications

Industry

Comparative analysis of

other industries

Industries

Creativity

FIGURE 10. E-BUSINESS QUALITATIVE CRITERIA APPLICATION MODEL. SOURCE: COMPILED BY THE AUTHORS

Dimension Composition Content Information Accuracy

Information Completeness Appropriate amount of data Concise Representation Understandability

Form Interface Structural Quality Information Accessibility Information Packaging Quality

Time History Maintenance Quality Information Currency

Information, data quality

Input: start of e-business

Output: high quality e-business



208

C. Model application possibilities and perspective

E-business qualitative criteria application model (see Figure 10) possibilities occurs when there is a wish to build as much as possible to the quality oriented e-business. Since there is no universally accepted e-business qualitative criteria, which can be used as a guide through the launch of e-business, this e-business qualitative criteria application model could be used for the development of quality-oriented e-business.

E-business qualitative criteria application model possibilities can be divided into two parts:

To rely on this model in early stages of e-business. Then the model becomes applicable from the input. When launching e-business, after forming an idea firstly it is important to answer to added value curve questions. Further there should be carried out customer satisfaction, the information, data quality analysis. Finally, it is important to determine e-business level of creativity. After all the analysis, it will be clear what to do, what requirements must be met in order to further develop a high quality e-business.

To rely on this model improving an existing e-business. Then e-business analysis based on all e-business qualitative criteria should be done in order to find the requirements that e-business does not meet and improve it. The highest quality will also be achieved when e-business will meet all e-business qualitative criteria requirements.

Any model perspective gets the answer to three basic questions: what is done? How is done? Who does? [2] (see Figure 11). When applying e-business qualitative criteria application model (whether it applies in early stages of e-businesses or in existing business in order to improve it), there should be answered the following questions.

FIGURE 11. MODELLING PERSPECTIVES. SOURCE: AYTULUN, GUNERI, 2008, P. 2746.



209

The most common answer to the question what is done? (this is the object) is what you need to improve. In e-business qualitative criteria application model case, this would be e-business qualitative criteria and its requirements, which analyzing e-business does not fit. The answer to the question of how is done? (this is an activity, that needs to be carried out) in this case it would be activities that should be carried out in order to meet e-business qualitative criteria and its requirements. The answer to the question who does? (this is the role of the person who will perform the work) in e-business qualitative criteria application model case this would be the person who will carry out the activities necessary to make in order to meet e-business qualitative criteria and its requirements.

In summary, it can be concluded, that e-business qualitative criteria application model possibilities reveal in two ways – when it is applied in early stages of e-businesses or in existing business in order to improve it. When analyzing e-business qualitative criteria application model perspectives there should be answered 3 basic questions: What is done? How is done? Who does? After answering to these questions, it becomes clear e-business qualitative criteria application object, activity and role.

V. CONCLUSION AND RECOMMENDATIONS

1. Theoretical aspects of e-business qualitative criteria include e-business qualitative criteria formation guidelines. After analyzing theoretical aspects of e-business quality, there were defined 4 e-business qualitative criteria: matching the value curve; orientation to the customer; information and data quality; creativity. It is important to focus attention on the customer's perspective, highlight the key requirements to ensure customer satisfaction. In terms of e-business quality, the quality of information is important, because it determines customer perception of goods or services quality. Creativity can be associated with e-business quality. In e-business case analysis, there occur 3 possibilities for creativity to reveal.

2. An expert qualitative opinion survey was carried out in which nine e-business experts were interviewed. Before analyzing the obtained data there was clarified expert’s opinions compatibility. The survey data analysis showed that in order to develop e-business and seek for the highest quality, there would be able to use the value curve and create e-business primarily answering four questions defined in this curve. Experts declare that in order to develop e-business and seek for the highest quality, there could be added more question to the value curve. Also, the survey data analysis showed that the most important e-business qualitative criteria is orientation to customer.

3. After proposing e-business qualitative criteria, there was created a model of their application. The model was created using the qualitative process descriptions. E-business qualitative criteria application model starts from the input – start of e-business. When launching e-business, after forming some idea, firstly it is important to answer to added value curve questions. Further there should be carried out customer satisfaction, the information, data quality analysis. Finally, it is important to determine e-business level of creativity. After analyzing business in accordance with all e-business qualitative criteria, it can be seen what needs to be improved, and the direction in which to do so, because the output (high-quality e-business) will be achieved only when e-business in great extent or completely satisfy these criteria. E-business qualitative criteria application model possibilities occur when there is a



210

wish to build as much as possible to the quality-oriented e-business. E-business qualitative criteria application model implementation possibilities can be divided into two parts: to rely on this model in early stages of e-business or rely on this model improving an existing e-business. Any model perspective gets the answer to three basic questions: what is done? How is done? Who does? The answers to these questions would help to clarify finally the e-business qualitative criteria application object, activity and role.

REFERENCES [1] Aburukbaa R., Masaud-Wahaishiab A., Ghenniwaa H., Shen W. (2009). Privacy-based Computation

Model in E-business. International Journal of Production Research, Vol. 47, No. 17, 4885–4906.

[2] Aytulun S. K., Guneri A. F. (2012). Business Process Modelling with Stochastic Networks. International Journal of Production Research, Vol. 46, No. 10, 2743–2764.

[3] Balaraman P., Kosalram K. (2012). E-Commerce Evaluation and E Business Trends. Information Engineering and Electronic Business, Vol. 5, 9-16.

[4] Baležentis A., Žalimaitė M. (2011). Expert Assessments for the Development of Innovations in the Analysis of Factors: Lithuanian Innovative Companies Assessment. Management theory and studies for rural business and infrastructure development, No. 3 (27), 23-31.

[5] Basha M. S. S., Dhavachelvan P. (2011). Survey on e-business Models. International Journal of Engineering Science and Technology, Vol. 3, No. 5, 3631 – 3639.

[6] Brown G. H. The E-Commerce Imperative, 2013. Retrieved from http://info.cassidyturley.com/blog/bid/294841/E-Commerce-Imperative-2013

[7] Chen S. (2001). Strategic Management of e-business. Wiley Editorial.

[8] Ding M. (2008). Perfect 10 v. Amazon.com: a Step toward Copyright’s Tort Law Roots. Berkeley Technology Law Journal, Vol. 23:373, 373–403.

[9] Donici A. N., Maha A., Ignat I., Maha L. G. (2012). E-Commerce across United States of America: Amazon.com. Economy Transdisciplinarity Cognition, Vol. 15, Issue 1, 252 – 258.

[10] Enders A., Jelassi T. (2009). Leveraging Multichannel Retailing: The Experience of Tesco.com. MIS Quarterly Executive, Vol. 8, No. 2, 89-100.

[11] Harris L., Dennis Ch. (2008). Marketing the e-Business. Routledge.

[12] Hsiao L., Chen Y. (2012). Returns Policy and Quality Risk in E-Business. Production and Operations Management, Vol. 21, No. 3, 489–503.

[13] Iqbal M., Khan A. A., Naseer Q. (2013). A Legal Perspective of e-businesses and e-marketing for Small and Medium Enterprices. International Journal of Managing Information Technology, Vol.5, No.1, 1–11.

[14] Jelassi T., Enders A. (2005). Strategies for e-business. Creating Value through Electronic and Mobile Commerce. Concepts and cases. Pearson Education Limited.



211

[15] Jin Kim Y., Kishore R., Sanders G. L. (2005). From DQ to EQ: Understanding Data Quality in the Context of e-business Systems. Communications of ACM, Vol. 48, No. 10, 75-81.

[16] Keblis M. F., Chen M. (2006). Improving Customer Service Operations at Amazon.com. Interfaces, Vol. 36, No. 5, 433–445.

[17] Kim W. Ch., Mauborgne R. (1999). Creating New Market Space. Harward Business Review, January – February, 83 – 93.

[18] Koronios A., Xu H. (2004-2005). Understanding Information Quality in e-business. Journal of Computer Information Systems, P. 73–82.

[19] Laudon K. C., Laudon J. P. (2007). Essentials of Business Information Systems. Seventh edition. Pearson Education.

[20] Libby R., Blashfield R. (1978). Performance of a Composite as a Function of a Number of Judges. Organizational Behavior and Human Performance, Vol. 21, 121-129.

[21] Milchev G. (2012). A Methodology for Managing the Effectiveness of Enterprise E-business Systems. Journal of the University of Chemical Technology and Metallurgy, Vol. 47, No. 2, 217-222.

[22] Mohamad R., Ismail N. A. (2013). The Extent of E-Business Usage and Perceived Cumulative Benefits: A Survey on Small and Medium-Sized Enterprises. Information Management and Business Review, Vol. 5, No. 1, 13-19.

[23] Mohanty R. P., Seth D., Mukadam S. (2007). Quality Dimensions of E-Commerce and their Implications. Total Quality Management, Vol. 18, No. 3, 219–247.

[24] Mudambi S. M., Schuff D. (2010). What Makes a Helpful Online Review? A Study of Costumer reviews on Amazon.com. MIS Quarterly, Vol. 34, No. 1, 185-200.

[25] Podvezko V. (2005). Experts Estimate Compatibility. Technological and Economic Development, Vol. 11, No. 2, 101- 107.

[26] Shetty V., Sarojadevi H. (2012). E-Business Performance Issues, Quality Metrics and Development Frameworks. International Journal of Computer Applications, Vol. 55, No.7, 42–47.

[27] Sidekerskienė T. Models Based System and Process Management. Retrieved from http://www.personalas.ktu.lt/~tatside/studijos/Simucio/TSidekerskiene_MPSPV_pataisytos.ppt

[28] Sun H. (2010). Transferring Attributes of E-Commerce Systems into Business Benefits: a Relashionship Quality Perspective. Journal of Electronic Commerce Research, Vol. 11, No 2, 92 – 109.

[29] Tidikis R. (2003). Social Science Research Methodology. Vilnius: Lithuanian Law University Publishing Center.

[30] Zilber S. N., de Araśjo J. B. (2012). Small Companies Innovations in Emerging Countries: E-Business Adoption and its Business Model. Journal of Technology Management&Innovation, Vol. 7, No. 2, 102-116.



212


Tadas Limba is a Head of Institute of Digital Technologies at Mykolas Romeris University in Vilnius, Lithuania. He got B. Sc. in Politics from Vilnius University, 1999 and B. Sc. in Law from Mykolas Romeris University, 2010. He got M. Sc. in Public Administration from Mykolas Romeris University, 2001 and M. Sc. in Business Law from Mykolas Romeris University, 2012. Tadas Limba also got his Ph. D. in Management and Administration from Mykolas Romeris University, 2009. Tadas Limba is an Associate Professor from 2010. Tadas Limba has published over 20 articles in Lithuanian and foreign scientific

journals, monograph, textbook, focused on e-government and e-business. His additional areas of research and expertise are – IT law regulation and policy; digital content, digital media, privacy and data protection issues. Tadas Limba is a member of Lithuanian Computer Society since 2007. Since 2013 he is Asia Center Board Member, South Korea's representative at Mykolas Romeris University. He is visiting professor at Zaragoza University in Spain. He plays an active role in international communication and development of joint double degree studies program with South Korea Dongseo University. Tadas Limba made presentations in international and national conferences. Tadas Limba is fluent in English, Spanish and Russian, he is also elementary user of German.

Gintarė Gulevičiūtė was born in Panevėžys, Lithuania in 1989. She got B. Sc. in Public Administration in 2008 and helds M. Sc. in Electronic Business Management from Mykolas Romeris University. Now she is an assistant of Institute of Digital Technologies at Mykolas Romeris University. In 2013 and 2014 she has published some papers – “Peculiarities of E-Government Services Implementation in European Union”; “Holistic Electronic Government Services Integration Model: from Theory to Practice” in Lithuanian and foreign scientific journals. Her areas of interest are e-government, e-business,

business communication and digital contents. Gintarė Gulevičiūtė is the coordinator of Digital Content Academy at Mykolas Romeris University. During her study years she has organized innovative conference “Future business 2013“ at the University. She plays an active role in international communication and development of joint double degree studies program with South Korea Dongseo University. Now she is also a coordinator of Summer School of Communication “Science and Art of Communication” for Chinese students at Mykolas Romeris University.


A UML PROFILE FOR USE CASES MULTI-INTERPRETATION

Authors

Mira Abboud LaMA Laboratory - Lebanese University, Tripoli, Lebanon.

[email protected] Tripoli, Lebanon

Hala Naja LaMA Laboratory - Lebanese University, Tripoli, Lebanon.

[email protected] Tripoli, Lebanon

Mohamad Dbouk Faculty of Science, Lebanese University, Beirut, Lebanon.

Bilal El Haj Ali Faculty of Science, Lebanese University, Beirut, Lebanon.

[email protected] Beirut, Lebanon

[email protected] Beirut, Lebanon

Abstract

Decomposition is a crucial activity adopted when analyzing and designing complex software intensive systems. It allows to describe a system as a set of less complex models dedicated to different system aspects. In this field, UML proposes 5 related Views that help to understand the architecture of the system. Each one focuses on a particular aspect of the system. In this paper, an additional decomposition capability based on actors is introduced. The proposed approach is illustrated by a case study.

Key Words

UML Views, Separation of concern, Personalization of users’ requirements, multi-interpreted use cases.

I. INTRODUCTION

Decomposition enables us to manage the complexity of software systems that we develop and to accomplish greater reuse and maintainability. Usually Views are used during software development aiming to decompose the whole software as the combination of partial artifacts. In this field, UML proposes 5 related Views that help to understand the architecture of the system. They are essential constituent of a system model. They collaborate to describe the system according to different perspectives. Each view focuses on a particular aspect of the system:

A UML profile for use cases multi-interpretation Mira Abboud, Hala Naja, Mohamad Dbouk and Bilal El Haj Ali


functional (exposing the requirements of the system), design (capturing the vocabulary and solution space), process (modeling processes), deployment (focusing on system engineering issues) or implementation (addressing the physical realization of the system). In each view, several models are produced.

When the system is complex dealing with a great number of actors and requirements, the decomposition into view models still enable to manage the large number of artifacts that are produced. In fact, all requirements realizations are tangled in the view models, making hard the understanding of the system functionalities according to actors. Therefore, in order to make easier the understanding of this big amount of artifacts, an additional decomposition capability is proposed. It is based on actors’ interests. The actor separation principle gathers the artifacts related to an actor goal, allowing to understand easily each actor requirements.

The paper is organized as follows: section 2 presents the study motivation. In sections 3 and 4 a brief review of related work and a background are provided. Section 5 introduces the case study and reveals the limitations of UML. Section 6 details the proposed extension of the UML metamodel for use cases multi-interpretation. Section 7 shows a meta-model instance extracted from the case study. Section 8 concludes the paper.

II. THE APPROACH MOTIVATION This study is based on UML which has become a standard language for models specification

and design in object-oriented software development. It deals with modeling complex systems dealing with a large number of actors and requirements. In such systems, modeling activity produces large number of artifacts. Hopefully, those artifacts are distributed among different View models making easier the understanding of the whole system model. But we assume that the View decomposition principle in UML is not sufficient for the two following reasons:

1. UML is claimed to be Use Case Driven [1]. That means use cases are used as a primary artifact for establishing the desired behavior of the system, for verifying and validating the system’s architecture, for testing, and for communicating among the stakeholders of the project. We assume that with the View decomposition principle, it becomes extremely difficult to trace actors’ goals and requirements (i.e. use cases) across the different Views. Indeed, with this decomposition principle, all requirements and their realizations (called scenarios in UML) are scattered across several View models. It becomes hard to understand a use case since it is divided into several artifacts existing in several views.

2. When systems deal with many stakeholders and actors, it becomes hard to browse the different view models, since each one holds a huge number of artifacts. Furthermore, one same requirement can have different interpretations - and then realizations - depending on the actor considering that requirement. In result, all requirements realizations are tangled together making hard the understanding which scenarios are related to which actor. So it would be simpler if requirements realizations were separated.



www.elvedit.com 216

FIGURE 1: DECOMPOSITION ACCORDING TO TWO SEPARATION PRINCIPLES.

For the two reasons exposed above, we propose a second separation principle (based on actors’ interests) that will be used conjointly with the Views separation principle [2]–[3]. The combination of those two separation principles will provide two dimensions in the system description: the first is needed for the architectural decomposition of the system, the second for actors based decomposition (Figure 1).

III. RELATED WORK Viewpoints or Views have a long history in software engineering and related fields. The

motivation for Views is separation of concerns and complexity reduction. Here we make a brief review of some approaches in the fields of software modeling and design, software architecture and software development.

Aspect-oriented development techniques [4]–[5]–[6], Architectural Description Languages (ADL) [7] and in object-oriented modeling techniques [8]–[9]–[10].

In modeling and design field, the most recent and widely available incarnation of multiple views modeling is the Unified Modelling Language (UML). They are introduced for Models complexity decomposition. In UML, views are considered at two: - In the latest version of UML (UML 2.4.1), 14 types of diagrams are proposed. Each one

focuses on a specific aspect of the system and has an implicit view. UML also provides extension mechanisms for allowing users to define additional diagrams types and thus indirectly new views.

- In reference [11], VUML profile is defined for viewbased modeling approach where the viewpoint is considered only on UML class diagrams. A composition model [10] is proposed to merge those models into a global one. In this approach, it is unclear how viewpoints are used during the modeling process. We assume that viewpoints have to be a first-class concept used all along the modeling process e.g. from requirement analysis to software development. We work in this direction.



In Software Development field, the Aspect-oriented development techniques consider that a number of software development concerns could not be handled using the modularization boundaries inherent in object-oriented languages and propose new artifacts (beyond method, class or package) to separate new kinds of concerns that tend to be amalgamated in object-oriented paradigms [6]. In this area, several methods are proposed : - The Subject-Oriented Programming (SOP) [4] technique addresses the functional

requirements. It views object oriented applications as the composition of several application slices representing separate functional domains. Each slice called subject consists of a self-contained, object-oriented program, with its own class hierarchy. Then, a composition language is defined allowing the composition of class hierarchies (subjects).

- The Aspect-Oriented programming (AOP) [5]–[12] technique addresses the non-functional requirements, including architectural aspects, error handling, security, distribution, persistence, etc. It defines aspect as an implementation of a concern that pertains to several objects in collaboration. An aspect is a code module addressing specific concern and that cross-cut various components in an application.

- The View-Oriented programming technique [13] considers each object of an application as a set of core functionalities available, directly or indirectly, to all users of the object, and a set of interfaces specific to particular uses, and which can be added or removed during run-time.

In software architecture, where the high-level structure of a software architecture is described, an architectural view is [14] a way of looking at an architecture. Each view may have a different concept of components and relationships. In fact: - A number of approaches are predefined on a set of fixed views. [15] presents an extensive set

of constructs called columns which are close to Views. The Reference Model for Open Distributed processing defines exactly five Architectural views. Also in UML [16], the system architecture is specified in 4+1 Views. The fifth view is used to validate the other views and their interactions.

- In Architecture Description Framework [17], views are not limited but can be specified according to the architect needs. Also in this approach, Views are constructive (i.e. not derived from a primary representation) and are governed by type-like entities called viewpoints.

IV. BACKGROUND UML is based on a four-layer architecture consisting of a meta-metamodel, a metamodel, a

user-defined model or design, and objects. The UML metamodel is an instance of the meta-metamodel defining the UML language. A user-defined analysis model or system design presented in the UML is an instance of the UML metamodel. Application-specific data is stored in objects created with classes specified in the design. In the subsequent paragraphs, we briefly present two notions deeply related with our approach: Views and profile mechanism.



www.elvedit.com 218

A. Views in UML

In UML notation, Kruchten [17] defines the following 5 views: The use case view of a system encompasses the use cases that describe the behavior of the

system as seen by is end users. With UML, the static aspects of this view are captured in use case diagrams; the dynamic aspects of this view are captured in interaction diagrams.

The design view encompasses the classes, interfaces and collaborations that form the vocabulary of the problem and its solution. This view supports the functional requirements. With UML, the static aspects are captured in class diagrams.

The process view encompasses the threads and processes that form the system’s concurrency and synchronization mechanisms.

The implementation view encompasses the components that are used to assemble and release the physical system.

The deployment view encompasses the nodes that form the system’s hardware topology on which the system executes. This view addresses the distribution, delivery and installation of the parts that make up the physical system.

B. Profile Mechanism in UML

In UML 2.0 [12], the UML metamodel can be customized while using Profile mechanism. A profile defines limited extensions to a reference metamodel with the purpose of adapting the metamodel to a specific platform or domain.

A stereotype defines how an existing metaclass may be extended and enables the use of a domain specific terminology in place or in addition to the ones used for the extended metaclass. It can generalize or specialize another stereotype and can have properties which may be referred to as tag definitions. When a stereotype is applied to a model element, the values of the property may be referred to as tagged values.

V. THE ONLINE AUCTION SYSTEM CASE STUDY In order to highlight the importance of the actor separation principle and to illustrate its

applicability, a case study based on an online auction system is introduced. This online system is dedicated for selling and buying vehicles. In this system, three main actors are identified as the following: Auctioneer: the person managing the selling and buying process. He/she is the system

administrator responsible for adding items to the wall, accepting or refusing bids, creating customer accounts, controlling the buying and selling process, following up payments and selling requests. Figure 2 shows the use case diagram according to the Auctioneer.

Public customer: a customer with no license who can only bid on public items and is prevented from many services. In fact, he/she cannot bid on all items and cannot put virtual bids (live bids). Figure 3 shows the use case diagram according to the Public customer.



Licensed customer: a customer with a trading license and having more permissions and services than the Public customer. This person can place bids on any item, make payments, send delivery requests and watch the salvage history and details for each item (some cars have salvage history showing if it was involved in accidents).The Licensed customer can also send requests to the Auctioneer to sell a car and then the Auctioneer may accept or refuse. Figure 3 shows the use case diagram according to the Licensed customer.

Guest: a user who can only browse the inventory. He/she does not have account in the system and cannot log in to the system. The guest can view vehicle photos listed in the auction; each photo is labeled by the name of the item without any further detail. Also a guest can view the website advertisement pages and the online help sections for registration and becoming a member.

As shown in Figures 2 and 3, the use case diagram shows several use cases (i.e. requirements). Some of them are shared between actors (those colored in light blue in the two figures).

FIGURE 3: USE CASE DIAGRAM ACCORDING TO THE LICENSED AND PUBLIC CUSTOMERS.

A. Use Cases with Different Interpretations

From use case diagrams in Figures 2 and 3, the following observations are made:

FIGURE 2: USE CASE DIAGRAM ACCORDING TO THE AUCTIONEER.



www.elvedit.com 220

The Auctioneer and Licensed customer share use case Add items to the wall. The Auctioneer, Guest, Public and Licensed customer share use case Browse Inventory. The Licensed and Public customer share use cases Place delivery request and Place bid.

Even though each use case mentioned above has the same goal, each one realizes its goal through a different sequence of messages (i.e. scenarios). In the subsequent paragraph, we develop this issue detailing 3 shared use cases among actors.

1. Use case Add items to the wall As its name indicates, use case Add items to the wall aims at adding vehicles to the wall.

Tables 1 and 2 detail two different main scenarios for this use case according to respectively the Licensed customer and Auctioneer. Even though the use case goal is the same, the main success scenarios differ because: An Auctioneer can insert any item and display it on the wall without any previous permission

from any user. See preconditions in Table 2. A Licensed customer needs to have the permission of the Auctioneer, in order to add an item on

the wall. In fact, the Licensed customer sends a request to the Auctioneer to add a set of items. The Auctioneer then sends back to the customer an acceptance code to be used when the customer inserts his items. See preconditions in Table 1.

This use case is written following Alistair Cockburn template and exposed in a table format in order to clarify the difference between both scenarios.

TABLE 1: MAIN SUCCESS SCENARIO OF ADD TABLE 2: MAIN SUCCESS SCENARIO OF ADD ITEMS TO THE WALL ACCORDING TO ITEMS TO THE WALL ACCORDING TO THE LICENSED CUSTOMER. THE AUCTIONEER.

Use Case name Add items to the wall. Primary actor Licensed customer. Preconditions The Licensed customer is logged in

the system.

A request is sent and an acceptance code is given.

The "add items" screen is launched. Success End

Condition An item is added to the wall.

Main Scenario Action 1 Customer fills the acceptance code. 2

The system checks the code correctness.

3 The auctioneer fills the required information for an item.

4 The System adds the item to the wall. 5 Repeat steps 1 and 2 until no more

items to be added.

Use Case name Add items to the wall. Primary actor Auctioneer. Preconditions The auctioneer is logged in to the

system. The "add items" screen is launched.

Success End Condition

Items are added to the wall.

Main Scenario Action 1 The Auctioneer fills the required

information for an item. 2 The system adds the item to the wall.

3 Repeat steps 1 and 2 until no more items to be added.



2. Use case Browse inventory This use case is presented to illustrate a base use case which is related by an inclusion

relationship to another use case and the include relationship holds only for specific actors (see figure 3). According to all actors, the main goal of Browse Inventory is to browse the items. However, the displayed information is different. In fact: Guest can only view the cars photos and their model numbers. Public customer can only view the basic information of the items consisting of the car photo,

brand, model, the current collision or status. Only Auctioneer and Licensed customer can see the history of items (previous accidents,

odometer, previous owners, etc.).

Therefore, depending on actors, Browse inventory maintains different include relationships (Figure 3). Indeed, for actors, it includes use case Browse basic information (for all actors except the Guest) and in other cases it includes use case Browse history (for only the Auctioneer and licensed customer).

3. Use case Place delivery request Place delivery request is presented to illustrate a base use case which is related by an extension

relationship to an addition use case and the extends relationship holds only for specific actors. It consists to place a request for delivering a vehicle and to suggest a suitable price range. This use case extends another use case (Browse delivery history) only for specific actor. In fact: The Licensed customer can consult the price history for delivery from same location for

previous items shipped before adding a suitable price range. The Public customer adds a suitable price range without having the possibility of consulting

previous shipping.

B. UML Limitation

Based on the above case study, we can make the following remarks: Even though a given use case is shared among different actors, it can have several

interpretations (i.e. main and variation scenarios); each interpretation is specific to a given actor. In UML, there is no way to cluster use case scenarios according to actors; this is a handicap for actors who are seeking for personalized requirements description.

Furthermore in UML, an include/extend relationship between use cases cannot be selective (i.e. cannot be considered for some actors and not for all of them). This matter of fact leads to ambiguity in use case diagrams interpretation.

In order to address the above problems encountered in the exposed case study, the following profile and extension of the UML metamodel has been proposed.



www.elvedit.com 222

VI. EXTENDING THE UML METAMODEL Our approach extends UML metamodel in two ways: First in defining a profile, called C-profile,

for providing a lightweight extension mechanism to the UML standard and second by introducing new classes and interfaces to the metamodel making possible the definition and manipulation of multi-interpreted use cases according to actors.

A. Profile for Use Cases Multi-Interpretations

1. Stereotypes for Viewpoints The UML metamodel is extended using the following stereotypes (Figure 4):

«viewPoint»: actor who needs to have a personalized access to his/her models. It changes the graphical appearance of an actor. The icon attached to this stereotype represents a stickman with a happy face replacing the empty circle traditionally used in UML notation.

«multiconcern»: classifier (class, use case, etc.) with multiple representations according to several actors. Association concern [*] links a multiconcern classifier to its concern classifiers.

«concern»: classifier (class, use case, etc.) with a representation of a multiconcern classifier according to a specific viewpoint.

«includesOnly»: include relationship between use cases holding only for some viewpoints. «extendsOnly»: extend relationship between use cases that holds only for some viewpoints.

B. Interfaces and Associations for Stereotypes

As stereotypes are a special kind of metaclasses they can realize interfaces. The extended metamodel contains the following interfaces and associations (Figure 5): mc-Interface: interface which encapsulates the behavior of a mullticoncern classifier, with

operations allowing adding, removing, enabling or disabling viewpoints. c-Interface: which encapsulates the behaviour of a concern classifier with operations allowing

the access to the corresponding multiconcern classifier.

FIGURE 4: NEW STEREOTYPES FOR VIEWPOINTS HANDLING.



VII. USE CASE STUDY REVISITED Applying the proposed profile to existing UML metamodel, enhances the diagrams semantics

and allows personalized access to different artifacts produced while UML modelling. Applying C-profile to the online auction case study presented in paragraph V, implies that (1) the actors are extended by viewpoints, (2) some use cases like Add items to the wall, Browse Inventory and place delivery Request are considered as multiconcern and (3) some includes and extends relationships become includesOnly and extendsOnly. Figure 6 presents a portion of the use case diagram showing the added semantics when applying C-profile.

FIGURE 6: PORTION OF THE USE CASE DIAGRAN WITH C-PROFILE.

Figure 7 shows a meta-model instance example based on the previous case study and on the extended metamodel defined in paragraph VI.A.1. In this figure, the following remarks can be made: Each actor is an instance of metaclass Actor. To each actor is applied an instance of stereotype

viewpoint. Use case Add items to wall is a multiconcern object instance of metaclass UseCase. It realizes

interface mc-Interface and is linked via qualified association Concern to two other concern use

FIGURE 5: EXTENDED METAMODEL FOR VIEWPOINTS HANDLING



www.elvedit.com 224

cases; each one realizes interface c-Interface and encapsulates the multiconcern use case behavior according to a specific viewpoint.

FIGURE 7: AN INSTANCE META-MODEL SPECIFICATION.

TABLE 3: INTERFACES OPERATIONS.

VIII. CONCLUSION In this paper, a UML profile, called C-profile is proposed: It suggests a second separation

principle based on actors, allowing functional requirements decomposition based on actors goals (i.e. concerns). This principle is to be used conjointly with the Views separation principle decomposing an application according to architectural considerations. This approach aims: to trace easily one actor goals across the different modeling steps (requirement analysis, design

or implementation), to minimize the impact of an actor requirement changes, to allow new actors to be easily added to existing actors,

Interface name

Operation Signature

Precondtion

Postcondition

viewpoint enable N/A activates the current viewpoint. viewpoint disable N/A deactivates the current viewpoint.

mc-Interface

addConcern(name)

name is an active viewpoint name.

a concern classifier, according to the given viewpoint name, is added to the list map of concern classifiers.

mc-Interface

removeConcern(name)

name is an active viewpoint name already attached to the current multiconcern classifier.

a concern classifier is removed from the list map of concern classifiers.

mc-Interface

concerns()

N/A

returns a map returns a map <viewpoint name, concern classifier> consisting of all the concern classifiers attached to the current multiconcern classifier.

mc-Interface

getConcern (name)

name is an active viewpoint name already attached to the current multiconcern classifier.

returns the concern classifier according to name.

c-Interface

getMultiConcern()

N/A

returns the multiconcern classifier corresponding to the current concern classifier.



to allow new behaviors to be added to multiconcern use cases without reconsidering existing ones,

to be compatible with an incrementally approach. This approach is at its beginning. It should be defined rigorously. In fact the extended meta-

model should be specified by OCL constraints. Also, we should study the impact of multiview use cases used in use case models on the other models (such as structural, interaction...) defined during design and implementation .

REFERENCES [1] OMG. Unified modeling language (omg uml) infrastructure. Specification Document formal/2010-05-

03, May 2005. http://www.omg.org/spec/UML/2.3/Infrastructure.

[2] Jacobson, I. e. a. The Unified Software Development Process. Addison Wesley, 1999.

[3] Larman, C. Applying UML and patterns. Prentice Hall, 2005.

[4] Ossher, H. and al. Subject-oriented composition rules. In OOPSLAS’95, pages 235–250, 1995.

[5] Kiczales, G. and All. Aspect-oriented programming. In ECOOP’97. Springler-Verlag, LNCS 1241, 1997.

[6] Mili, H., Elkharraz, and A. Mcheick, H. Concerned about separation. In the 2004 International Workshop on Early aspects, AOSD’04, pages 211–221. Prentice-Hall, 2004.

[7] Gacemi, A., Seriai, A., and Mourad Chabane, O. Code generation from Architectural Multi-views Description. Journal of Object Technology, JOT, 4(3), 2005.

[8] Naja, H. La représentation multiple d’objets pour l’ingénierie. L’objet– logiciel, bases de donnèes, réseaux., 4(2), 1998.

[9] Naja, H. Multiview databases for building modelling. Automation In Construction., 8, 1999.

[10] Anwar, A., Ebersold, S., Coulette, B., Nassar, M., and Kriouile, A. A Rule-Driven Approach for composing Viewpoint-oriented Models. Journal of Object Technology, JOT, 9(2), 2010.

[11] Nassar, M., El Asri, B., Coulette, B., and Kriouille, A. Vers un profil uml pour la conception de composants multivues. 11(4), 2005.

[12] Majumdar, D. and Swapan, B. Aspect Oriented Requirement Engineering: A Theme Based Vector Orientation Model. Journal of Computer Science, InfoComp. http://www.infocomp. ETH Zurich, 2010.

[13] Mili, H. and al. View programming: Towards a framework for decentralized development and execution of oo programs. In Proc. of TOOLS USA’ 99, pages 211–221. Prentice Hall, 1999.

[14] Group, I.A.W. Ieee recommended practice for architectural description of software-intensive systems. Technical report, Institute of Electrical and Electronics Engineers, NY, USA (September 2000).

[15] Sowa, J., Zachman, J. Extending and formalizing the framework for information systems architecture. IBM systems journal 31(3), pages 590-616, 1992.

[16] Kruchten, P. The 4+1 view model of architecture. Software, IEEE 12(6) (1995) 42-50.



www.elvedit.com 226

[17] Rozanski, N., Woods, E. Software Systems Architecture: Working with Stakeholders Using Viewpoints and Perspectives. Addison-Wesley (Oct 2011).


A Grid-enabled Application for the Simulation of Plant Tissue Culture Experiments

Authors

Florence I. Akaneme Faculty of Biological Sciences, Plant Science and Biotechnology Dept. University of Nigeria

[email protected] Nsukka,, Nigeria

Collins N. Udanor Faculty of Physical of Physical Sciences, Computer Science Dept, University of Nigeria


Jane Nwachukwu

Faculty of Biological Sciences , Plant Science and Biotechnology Dept. University of Nigeria


Chibuike Ugwuoke

Faculty of Physical of Physical Sciences, Computer Science Dept, University of Nigeria


Carl .E.A Okezie Faculty of Biological Sciences , Plant Science and Biotechnology Dept. University of Nigeria


Benjamin Ogwo. Department of Technical Vocational Education, New York State University

[email protected] New York, USA.

Abstract

Plant Tissue culture is a method for plant propagation under in vitro conditions. Different types and parts of plants (known as explants) may be cultivated in vitro. These may be organs (roots, stems, shoot tips, leaves and fruit); tissues; cells (suspension cultures) and special tissues and organs such as embryos, anthers, pollen and protoplasts. Plant tissue culture is a laborious and time-consuming technique. Potentially, modeling or computer simulation can provide a useful method for gaining insight into these complex processes by reducing the time needed to screen numerous hormonal combinations. We present a simulation application based on multiple regression models deployed on a grid computing infrastructure. The application simulates the plant tissue culture experiments and predicts the amount and combinations of auxins and cytokinins needed to yield optimal growth of propagules. The results obtained from the simulation showed over 67% prediction accuracy as compared to the laboratory experiments. Key Words

Explants, In Vitro, Micropropagation, Plant Tissue Culture, Simulation.



www.elvedit.com 228

I. INTRODUCTION This collaborative research work was necessitated by the project: ‘Sustaining the Research and

Grid Computing Components of the University of Nigeria’s UNESCO – HP Brain Gain Project’ awarded to the University of Nigeria through a competitive bid process under the UNESCO-Hewlett Packard Brain Gain Initiative (UNESCO-HP BGI). As a follow-up to the ongoing plant tissue culture research activities, the project developed a simulation application to predict the outcome of various combinations of plant growth hormones. This was aimed at reducing the number of laboratory trials, time and the cost of the experiment. The application is deployed on the University of Nigeria Grid Computing infrastructure. A comparative analysis of this application and the results of the laboratory work are presented in this paper. The rest of the paper is organized as follows. Section II presents a brief review of related works. Section III shows details of the laboratory work, as well as the design of the software with a brief description of linear regression model on which this software is based. And section IV shows the results from the comparison of the software and experimental data, as well as sample output of the application.

II. REVIEW OF RELATED WORKS

Grid computing is a form of distributed and parallel computing, whereby a 'super and virtual computer' is composed of a cluster of networked, loosely coupled computers acting in concert to perform very large tasks [1]. Foster and Kesselman [2], defined a computational grid as a hardware and software infrastructure that provides dependable, consistent, pervasive, and inexpensive access to high-end computational capabilities.” Thanks to the growth of the Internet and high speed data networks, geographically distributed resources, such as storage devices, data sources, and supercomputers, are interconnected and can now be exploited by users around the world as single, unified resource. Apart from hardware resources, application software programs can now be shared. The grid has thus drastically reduced computing cost, as well as made scare resources available to researchers who were technologically disadvantaged a few years back. One of the most common grid applications is simulation application. Simulation is widely used for modeling real world processes in many different application areas, including manufacturing, construction, and computer science. It provides the study of various issues, such as feasibility, behaviour and performance without building the actual system, thus saving time, cost and effort [3], [4]. It is based on the applicability of the grid to many fields of research that we developed a simulation application to model the processes involved in plant tissue culture experiments.

Plant tissue culture is still in its empirical stage, involving a lot of trials and error. The experiment is time and material intensive, running into several months of laboratory efforts in trying to build hormonal combinations that will be best for mass propagation of a particular species. According to [5], the most variable or critical factors in plant tissue culture media are growth regulators or hormones especially auxins and cytokinins which are usually used in various combinations that can run into hundreds. The growth regulators are important in determining the developmental pathway of plant cells. Modeling or computer simulation will readily be of great help in reducing the time needed to screen the numerous hormonal combinations.

A Grid-enabled Application for the Simulation of Plant Tissue Culture Experiments Florence I. Akaneme, Collins N. Udanor, Jane Nwachukwu, Chibuike Ugwuoke, Carl .E.A Okezie and Benjamin Ogwo


Thus due to the potentials of plant tissue culture technique, innovative approaches to reduce labour requirements and costs are being developed.

According to Afreen [6], using machines to accomplish the various steps of micropropagation will help to cut down the production costs. Sluis [7], however, was of the opinion that automation of micropropagation work is not technologically simple and also not readily achievable economically. He further noted that the human eye-hand-brain combination is both highly sophisticated, technologically and incredibly inexpensive when considered on a global scale. Warren [8] had earlier reported that human operators are proving difficult to supersede because much judgment is required concerning the best tissue to transfer and the optimum timing of the various steps.

‘Methods that could be used for the routine propagation of all kinds of trees have not yet been developed, despite much research [9]. This fact has been reiterated recently by [10] who reported that cultural requirements for the process of plant tissue culture differ from species to species. The most appropriate conditions for a given species must always be evolved out of experimentation.

Bhojwani and Razdan [11], also wrote that the formulation of a suitable medium for an untested species, would naturally start with a well-known basal medium such as Murashige and Skoog (MS) [12]. Furthermore, they noted that by making minor qualitative or quantitative changes through a series of experiments, a new medium may be evolved to accommodate specific requirements of the plant material in question.

Prasad and Gupta [13] described the various applications of artificial neural networks (ANN) in in vitro plant culture systems. They observed that ANN can play central role as highly potential predictive modeling tool in in vitro plant culture studies. They further reported that neural computing offers reliable and realistic approach for describing in vitro culture of plant species even with minimal available information. The successes obtained after applying neural network technology have been phenomenal with a relatively modest experimental effort while consuming minimum amount of time. ANN based prediction of the behaviour of the in vitro derived plants in terms of their ex vitro survival rate and their rooting or organogenic ability could also be useful in large scale propagation.

III. MATERIALS AND METHODS

In this section, we present the laboratory experiment design and execution. We shall also briefly discuss the software design methodology.

A. Design of Experiments

The laboratory experiments were carried out at the Plant Tissue Culture laboratory of National Root Crops Research Institute, Umudike, Umuahia, Abia State, Nigeria. Shoot tip explants were excised from aseptically germinated buds of cocoindia on basal MS media. Multiple shoot induction from these explants were investigated on two culture media which were: Schenk and



www.elvedit.com 230

Hildebrandt (SH) [14]; Arnold and Eriksson (AE) [15]. To these two respective basal media were added, 30g of sucrose, 10mg/l L-cysteine, 100mg/l myo-inositol, and vitamins.

Different concentrations of an auxin, Naphthalene acetic acid (NAA) and cytokinin, 6-Benzyl amino purine (BAP) and 6-furfuryl amino purine (Kinetin) were also added to each of the media. The concentrations were, 0.0, 0.05, 0.1, 0.5, and 1.0mg/l of NAA and 0.0, 2.0, 4.0, 6.0, 8.0mg/l of BAP and Kinetin respectively. NAA was combined in all possible combinations with BAP to give 25 treatments and likewise NAA plus Kinetin. Therefore, a total of 50 treatment combinations were obtained for each media.

Each treatment combination was replicated 10 times thus giving a total of 500 culture vessels for each media. Thirty milliliters of the respective media were dispensed into each culture vessel. Three shoot tip explants were seeded into each culture vessel thus giving a total of 1500 explants per medium; however, data analysis was performed with the mean of the three with respect to the attributes in question. The attributes studied include: number of shoots, number of leaves, number of roots and plant height. Thus for NAA x Kinetin, 500 pieces of data were obtained for each of the attributes.

But for NAA x Kinetin, 250 pieces of data (corresponding to 250 culture vessels) were obtained for the attributes; number of shoots and number of leaves. The culture vessels were sealed with paraffin and aluminum foil and placed on shelves in a growth room. The vessels were exposed to a 16 hour photoperiod which was provided by white fluorescent tubes. The temperature in the growth room was maintained at 28+ 2oC by air conditioning units. A separate rooting stage media were not prepared because the plantlets rooted while still in the respective multiplication media.

The whole experiment lasted for twelve months. The first four months were used to generate the required number of shoot tips while the last eight months were used to screen the hormonal combinations for their effects on multiple shoot induction from the shoot tip explants.

B. Software design

The system design and implementation for the simulation application is presented in this section. We review briefly the theory of linear regression which the application uses to estimate the yield of the propagules. A matrix formulation of the multiple regression model is shown in the equation below.

In the multiple regression setting, it is more efficient to use matrices to define the regression model and the subsequent analyses. This is because of the potentially large number of predictors.

Here, we review basic matrix algebra, as well as learn some of the more important multiple regression formulas in matrix form. Starting with the simple case first, consider the following simple linear regression function:



γi = β0 + β1χ1i + β2 χ2i + εi for i = 1,... , n (i)

Let us assume i = 1 to n, we then obtain equations as below:

γ1 = β0 + β1χ11 + β2 χ21 + ε1 (ii) γ2 = β0 + β1χ12 + β2 χ22 + ε2 (iii) γn = β0 + β1χ1n + β2 χ2n + εn (iv)

At a glance, one would easily make out that the equations would be cumbersome to handle in

cases of numerous trials (where n is large), and we would also agree that a pattern seems to be observed. With the help of matrix, we can translate the following linear regression functions into a matrix notation:

(v)

One can now see that the matrix notation still gives us a simple linear regression that can be

expressed as:

γ = χβ + ε (vi) We can now express the multiple n equations in a simple linear form, and the linear regression

function now becomes even shorter and simpler, as shown in (v).

Now γ is the response variable in equation (v), while χ is the control variable and in this case we have two variables (χ1 and χ2) representing auxin and cytokinin concentrations respectively, the parameters to be estimated from the data, including the constant of interception β. These constants of interception are obtained by computation of the matrix, using the already known parameters (y for expected result, X1 for auxin concentration, and X2 for cytokinin concentration, and ε is the random error variable which is assumed to be zero. Once the constant of interception has been successfully calculated, over a wide range of trials, we could now trust the constant to help in the estimation of response values (yield), when other concentration of auxin and cytokinin are introduced.

In the implementation of this work, we used python programming language because of its ability to handle the manipulation of data, even at large scale. This gives us the ability to efficiently perform the matrix operations required in less time. With the help of python libraries such as numpy and scipy, the matrix operations were handled and results confirmed to be accurate.

From the model equation in (v) above, we used the existing data obtained from real case experiment to obtain our β values and assuming ε to be 0 at all times in equation (vi), we could easily predict the outcome of the yield of propagules (ߛ) with a given combination of growth



www.elvedit.com 232

hormones ((χ1 and χ2).

At the time of this report, we had 750 test cases, and this helps in the estimation of the β values, also giving room for more experimental values to be added, which would increase the accuracy of the predictions.

IV. RESULTS AND DISCUSSIONS In this section, we use statistical analysis tools to compare the results obtained from the

experimental data with that obtained from the software simulations. Regression graphs and tables were used to find the coefficients of determination (R2) between the experimental and simulation data. These graphs are shown in figures 1 – 12.

FIGURE 1 NUMBER OF SHOOTS IN AE MEDIUM (NAA X KINETIN)

FIGURE 2 NUMBER OF SHOOTS IN SH MEDIUM (NAA X KINETIN)

y = 1x - 2E-05R² = 0.4039

0

0.5

1

1.5

2

2.5

3

3.5

4

0 1 2 3 4

Lab

orat

ory

Software

y = 1x - 0.0001R² = 0.2049

00.5

11.5

22.5

33.5

44.5

5

0 1 2 3 4 5

Lab

orat

ory

Software



FIGURE 3 NUMBER OF LEAVES IN AE MEDIUM (NAA X KINETIN)

FIGURE 4 NUMBER OF LEAVES IN SH MEDIUM (NAA X KINETIN)

y = 0.6901x + 0.8424R² = 0.0083

0

1

2

3

4

5

6

7

5.2 5.3 5.4 5.5 5.6 5.7 5.8

Lab

orat

ory

Software

y = 1x + 2E-05R² = 0.2098

0

1

2

3

4

5

6

7

0 1 2 3 4 5 6

Lab

orat

ory

Software



www.elvedit.com 234

FIGURE 5 PLANT HEIGHT IN AE MEDIUM (NAA X KINETIN)

FIGURE 6 PLANT HEIGHT IN SH MEDIUM (NAA X KINETIN)

y = 0.9323x + 0.4077R² = 0.2661

0

1

2

3

4

5

6

7

8

9

0 2 4 6 8

Lab

orat

ory

Software

y = 1.0049x - 0.0518R² = 0.1938

0

2

4

6

8

10

12

14

16

0 2 4 6 8 10 12

Lab

orat

ory

Software



FIGURE 7 NUMBER OF ROOTS IN AE MEDIUM (NAA X KINETIN)

FIGURE 8 NUMBER OF ROOTS IN SH MEDIUM (NAA X KINETIN)

y = 1x + 1E-06R² = 0.614

0

5

10

15

20

25

0 5 10 15 20

Lab

orat

ory

Software

y = 1x - 2E-06R² = 0.2814

0

5

10

15

20

25

0 5 10 15 20

Lab

orat

ory

Software



www.elvedit.com 236

FIGURE 9 NUMBER OF SHOOTS IN AE MEDIUM (NAA X BAP)

FIGURE 10 NUMBER OF SHOOTS IN SH MEDIUM (NAA X BAP)

y = 0.9964x + 0.015R² = 0.1672

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0 0.5 1 1.5 2 2.5 3 3.5

Lab

orat

ory

Software

y = 0.9659x + 0.1075R² = 0.0236

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

3.1 3.15 3.2 3.25 3.3

Lab

orat

ory

Software



FIGURE 11 NUMBER OF LEAVES IN AE MEDIUM (NAA X BAP)

FIGURE 12 NUMBER OF LEAVES IN SH MEDIUM (NAA X BAP)

Figures 1 to 12 are graphical explanations of variations in the attributes studied in the laboratory as predicted by the software. The slopes of the regression lines in figures 3, 10, 11 and 12 are parallel to the X-axis, thus there are no relationships between the laboratory and the software data. On the other hand, the slopes of the other figures depict some linear relationships between the two sets of data.

The coefficients of determination (R2) authenticate the above results. The R2 values for figures 3, 10, 11 and 12 were 0.008 (0.8%), 0.023 (2.3%), 0.000 (0%) and 0.004 (0.4%) respectively. It

y = 1.0443x - 0.2402R² = 0.0005

0

1

2

3

4

5

6

7

8

5.3 5.32 5.34 5.36 5.38 5.4

Lab

orat

ory

Software

y = 0.9998x + 0.0006R² = 0.0046

0

0.5

1

1.5

2

2.5

3

3.5

4

2.94 2.96 2.98 3 3.02 3.04

Lab

orat

ory

Software



www.elvedit.com 238

follows that some 98%, 97.7%, 100%, and 96% respectively of the variations in laboratory data shown in those figures are not accounted for by the software. In contrast, 61.4% of the variations in the number of roots obtained in the laboratory on AE medium + NAA + Kinetin were predicted by the software.

TABLE 1: ESTIMATES OF LINEAR REGRESSION COEFFICIENTS FOR ATTRIBUTES STUDIED

Attributs MS VR AE(NAA x Kinetin) No. of shoots

Regression Residual

5.005 0.321

15.585***

SH(NAA x Kinetin) No. of shoots

Regression Residual

2.254 0.380

5.926*

AE(NAA x Kinetin) No. of leaves

Regression Residual

0.096 0.500

0.192 NS

SH(NAA x Kinetin) No. of leaves

Regression Residual

3.611 0.591

6.108*

AE(NAA x Kinetin) Plant height

Regression Residual

10.189 1.221

8.341**

SH(NAA x Kinetin) Plant height

Regression Residual

17.705 3.203

5.528*

AE(NAA x Kinetin) No. of roots

Regression Residual

157.036 4.292

36.585***

SH(NAA x Kinetin) No of roots

Regression Residual

33.437 3.712

9.009**

AE(NAA x BAP) No. of shoots

Regression Residual

3.248 0.703

4.619*

SH(NAA x BAP) No. of shoots

Regression Residual

0.050 0.090

0.555 NS

`AE(NAA x BAP) No. of leaves

Regression Residual

0.017 1.496

0.012 NS

SH(NAA x BAP) No. of leaves

Regression Residual

0.009 0.082

0.106 NS

NOTE: * = significant at 5% level of probability, ** = significant at 1% level of probability, *** = significant at 0.1% level of probability, NS = not significant.

The regression mean squares in Table 1 above show the amount of total variation in the laboratory data that can be explained by the software model. The effectiveness of the model was greatest in AE and SH media supplemented with NAA and Kinetin for the following attributes in that order – number of roots per explant and plant height. The error mean square shows the amount of variation in the laboratory data that are left unexplained by the model and this was found to be highest in the AE and SH media supplemented with NAA and Kinetin for the number of roots per explants.



The table also shows that there were no significant relationships between the laboratory and software data for the following attributes: No. of leaves for AE medium supplemented with NAA x Kinetin, No. of leaves for AE and SH media supplemented with NAA and BAP, and No. of shoots for SH media supplemented with NAA and BAP. The rest showed some linear relationships and the maximum was observed with AE medium + NAA + Kinetin on number of roots.

V. CONCLUSION Out of 12 tests conducted as indicated in Table 1, only 4 results did not give signification

coefficient of determination, while 8 were significant. This means that the software has 66.67% overall ability to predict the outcome of the laboratory trials. Future work will focus on the development of a newer version of the software that will increase the prediction accuracy up to 90% and above.

APPENDIX In this section, we show some sample screen shots from Software.

Appendix A: THE HOME PAGE



www.elvedit.com 240

Appendix B: FILE UPLOAD PAGE

Appendix C: FILE LIST PAGE

Appendix D: APPLICATION PAGE



Appendix E: THE RESULT PAGE

ACKNOWLEDGMENT We acknowledge the support from the United Nations Education Scientific and Cultural Organization (UNESCO), for providing the funds for this research project and Hewlett Packard (HP) for providing the equipment for the deployment of the grid infrastructure. And also University of Nigeria for providing the enabling ground and other forms of support.

REFERENCES [1] Wikipedia, (2013) “Grid Computing”, en.wikipedia.org/wiki/Grid_computing. accessed on 26th

September, 2013 [2] Foster, I and Kesselman, C (1998) “The Grid: Blueprint for a New Computing Infrastructure”, Morgan

Kaufmann Publishers, San Francisco, CA, USA, 677pp. [3] Senin, R. Groppetti, A. Rossi and D. Wallace. Integrating Manufacturing Simulation Tools Using Distributed Object Technology. 4th IEEE/IFIP International Conference on Information Technology for Balanced Automation Systems in Production and Transportation, Berlin, September 2000. [4] Suri, R and Tomsicek, M Rapid Modeling Tools for Manufacturing Simulation and Analysis.

Proceedings of the 20th Conference on Winter Simulation, San Diego, 1988. [5] Pierik, R. L. M (1997) In vitro culture of higher plants, Kluwer Academic Publishers, Dordrecht, 348

pgs [6] Afreen F. (2006) “Temporary immersion bioreactors” In: Gupta S. D. and Ibaraki, Y (eds.) Plant Tissue

Culture Engineering, Springer, Netherlands, pp 187 – 201. [7] Sluis, C. J. (2006) “Integrating automation techniques with commercial micropropagation” In: Gupta S.

D. and Ibaraki, Y (eds.) Plant Tissue Culture Engineering, Springer, Netherlands, pp 231 – 251.



www.elvedit.com 242

[8] Warren, G (1991) The regeneration of plants from cultured cells and tissues In: Stafford, A and Warren, G (eds.), Plant Cell and Tissue Culture, John Wiley and Sons, Chichester, pp. 82 – 100.

[9] George, E. G and Sherrington, P. D (1984) Plant propagation by tissue culture: Handbook and Directory of Commercial Operations, Exegetics Ltd, Basingstoke, 709pp.

[10] Sathyanarayanan, (2010) “Understanding Plant Tissue Culture” www.2indya.com/2010/03/19/business-in-plant-tissue-culture/.../1/. Accessed on 10th October, 2010.

[11] Bhojwani, S. S and Razdan, M. K. (1983) Plant Tissue Culture: Theory and Practice, Developments in Crop Science (5), Elsevier Science Publishers, New York, 502pp.

[12] Murashige, T and Skoog, F (1962) A revised medium for rapid growth and bioassays with tobacco tissue cultures. Physiol Plant, 15: 473 – 497j

[13] Prasad, V. S. S and Gupta, S. D (2006) “Applications and potentials of artificial neural networks in plant tissue culture” In: Gupta S. D. and Ibaraki, Y (eds.) Plant Tissue Culture Engineering, Springer, Netherlands, pp47 – 67.

[14] Schenck, R.U and Hildebrandt, A. C (1977) Medium and Techniques for Induction and Growth of Monocotyledonous and Dicotyledonous Plant Cell Cultures, Can. J. Bot, 50: 199 – 204.

[15] Arnold, S and Eriksson, T (1977) Induction of adventitious buds on embryos of Norway Spruce grown in vitro Physiologia plantarum, 44: 283 – 287.


Clustering Evolutionary Computation for Solving Travelling Salesman Problems

Authors

Tanasanee Phienthrakul Department of Computer Engineering, Faculty of Engineering, Mahidol University

[email protected] NakornPathom, 73170, THAILAND

Abstract

This paper proposes the methods for solving the traveling salesman problems using clustering techniques and evolutionary methods. Gaussian mixer model and K-means clustering are two clustering techniques that are considered in this paper. The traveling salesman problems are clustered in order to group the nearest nodes in the problems. Then, the evolutionary methods are applied to each cluster. The results of genetic algorithm and ant colony optimization are compared. In the last steps, a cluster connection method is proposed to find the optimal path between any two clusters. These methods are implemented and tested on the benchmark datasets. The results are compared in terms of the average minimum tour length and the average computational time. These results show that the clustering techniques are able to improve the efficiency of evolutionary methods on traveling salesman problems. Moreover, the proposed methods can be applied to other problems.

Key Words

Evolutionary Computation, Gaussian Mixer Models, K-means Clustering, Traveling Salesman Problems

I. INTRODUCTIONTraveling salesman problem (TSP) is one of the most widely discussed problems in the

combinatorial optimization researches. This problem has been addressed extensively by mathematicians and computer scientists [1]. The applications of TSP are appeared in planning, scheduling, and searching in scientific and engineering fields, such as vehicle routing, manufacturing, computer operations, integrated circuit testing, cytological testing, and examination timetabling. In the simplest TSP, an optimal path of the shortest length is found on an undirected graph that represents cities or nodes to be visited [1]. The salesman in TSP will



www.elvedit.com 244

start at a node, visit all other nodes successively only one time each, and finally return to the starting node [1]. This problem is hardly solvable through finding the exact solution directly. Thus, many heuristics and approximation algorithms have been proposed to produce the useable solutions for TSP, such as neural networks [2], [3], simulated annealing [4], genetic algorithm [5],[6], and ant colony optimization [7]-[9] .

Both genetic algorithm and ant colony optimization are the most popular evolutionary algorithms with the diverse ranges of optimization applications. New techniques in genetic algorithm have been developed to solve TSP. In a research of Yang et al. [10], the generalized chromosome in genetic algorithm was proposed. They showed the method to create chromosomes for solving TSP. Liu and Zeng [11] applied reinforcement learning to genetic algorithm. They used this method to solve TSP. Albayrak and Allahverdi [12] solved TSP by genetic algorithm using a new mutation operator. These methods can produce the good solutions of TSP, but they still use a lot of computational time in the large TSP.

To improve the computational time, the concept of clustering is considered. If the nodes in the large TSP are clustered, the TSP problem should be smaller. The computational time of evolutionary computation for solving the smaller TSPs might be reduced. The similar concept can be found in the clustered traveling salesman problem (CTSP) researches [13]. The CTSP consists of determining a least cost Hamiltonian tour or cycle on graph in which the vertices of any clusters are contiguous [13]. There are some researches on CTSP, such as in [14], Anily, Bramel, and Hertz adjusted Christofides’ algorithm to get the shortest Hamiltonian path in each cluster and present a 5/3-approximation algorithm for the ordered CTSP. Guttmann-Beck et al., [15] developed the approximation algorithms that guarantee the performance of the ordered CTSP. Ding, Cheng, and He [16] used genetic algorithm to find the shortest Hamiltonian cycle for each cluster and developed genetic algorithm for ordered CTSP. However, the genetic algorithm technique for ordering the CTSP used a lot of computational time. Moreover, this technique did not always generate the good result.

In this paper, clustering algorithms are used to improve the efficiency of evolutionary computation algorithms in TSP. K-means clustering and Gaussian mixer models are compared. Genetic algorithm and ant colony optimization are used as 2 evolutionary computation algorithms in this research. The results are compared in the average minimum tour length and the average computational time. Moreover, this paper proposes a technique for clustered TSP connecting. The basic concepts of clustering and evolutionary computation are briefly reviewed in section II. Section III presents the clustering evolutionary computation algorithm that is the main idea of this paper. The experimentation results are showed in section IV and the conclusion is presented in section V.

II. BACKGROUND Clustering techniques and evolutionary computation are applied to solve the TSPs. K-means

and Gaussian mixer model will be compared in the clustering step. Genetic algorithm and ant colony optimization are used as the standard methods for solving the TSPs. The basic concepts of these algorithms are briefly reviewed in this section.

Clustering Evolutionary Computation for Solving Traveling Salesman Problems Tanasanee Phienthrakul


A. Clustering Techniques

A cluster is a set of objects in which each object is closer or more similar to every other object in the cluster than to any object not in the cluster. Clustering is a main task of explorative data mining, and a common technique for statistical data analysis. Clustering techniques are used in many fields, including machine learning, pattern recognition, image analysis, and information retrieval [17]. The aim of clustering is finding the useful groups of objects (clusters), where usefulness is defined by goals of the data analysis. For two dimensional data, there are many kinds of clustering techniques, such as prototype-based clustering or centre-based clustering, graph-based clustering, and density-based clustering [17]. In TSPs, the minimum tour length is the main objective of the path finding. Set of nodes or cities that are closer should be grouped to the same cluster. The centre of the cluster can be used as the prototype. Thus, K-means clustering and Gaussian mixture model which are prototype-based clustering will be considered.

1. K-means Clustering: Each object in K-means clustering will be assigned to precisely one of a set of clusters [18]. This method of clustering is started by deciding how many clusters, denoted by K. The value of K is generally a small integer. Then, K nodes are selected to be the initial centroids. These nodes can be selected in any way, but this method may be better if K nodes are chosen fairly far apart. There are many ways in which K clusters might potentially be formed. A method to measure the quality of a set of clusters is the sum of squares of the distances between each point and the centroid of its cluster. This value should be as small as possible.

All nodes will be assigned to a cluster by selecting a centroid that gives the minimum distance between node and centroid. Then, there are K clusters based on the original K centroids but the centroids may not be the true centroids of the clusters. Hence, the centroids of each cluster must be recalculated. Each node will be reassigned to the cluster with the nearest centroid. These processes are repeated until the centroids are no longer move. The K-means clustering algorithm and example are showed in Figure I.

K-means Algorithm

1. Choose an integer K. 2. Select K objects in an arbitrary fashion.

Use these as the initial set of k centroids. 3. Assign each of the objects to the cluster for

which it is nearest to the centroid. 4. Recalculate the centroids of the K clusters. 5. Repeat steps 3 and 4 until the centroids no

longer move.

FIGURE I: K-MEANS CLUSTERING ALGORITHM AND EXAMPLE.

2. Gaussian Mixer Models: A Gaussian mixer model is a collection of K Gaussian distribution. Each distribution represents a cluster of data points. Gaussian mixer model uses the expectation and maximization (EM) algorithm [19] to fit the Gaussian distributions to the data. The



www.elvedit.com 246

algorithm is started by defining the number of cluster (K) and chooses the parameter of K Gaussian distributions:

휆 = (휇 , 휇 , … , 휇 ,휎 ,휎 , … ,휎 ), (1)

when each cluster is the normal distribution with 푁(휇 ,휎 ).

The EM algorithm calculates the probability that each point belongs to each distribution and then uses these probabilities to compute a new estimate for the parameters. This iteration continues until the estimates of the parameters either do not change or change very little. The EM algorithm for Gaussian mixer model and the example are showed in Figure II.

EM Algorithm for Gaussian Mixer Models

1. Choose an integer K. 2. Randomly estimate the parameters of K

distributions. 3. Expectation Step: Calculate the probability

that each object belongs to each distribution. 4. Maximization Step: Use the probability to

find the new estimates of the parameters that maximize the expected likelihood.

5. Repeat step 3 and 4 until the parameters do not change or the parameters is below a specified threshold.

FIGURE II: EM ALGORITHM FOR GAUSSIAN MIXER MODEL AND EXAMPLE.

B. Evolutionary Computation

Evolutionary computation uses iterative progress, such as growth or development in a population [20]. The population is selected in a guided random search using parallel processing to achieve the desired end. Such processes are often inspired by biological mechanisms of evolution. Evolutionary computation techniques mostly involve meta-heuristic optimization algorithms such as evolutionary algorithms (comprising of genetic algorithms, evolutionary programming, evolution strategy, genetic programming, learning classifier systems, and differential evolution) and swarm intelligence (comprising of ant colony optimization and particle swarm optimization) [20]. Moreover, there are the lesser extent algorithms, such as self-organization such as self-organizing maps, growing neural gas, artificial life (digital organism), cultural algorithms, harmony search algorithm, and artificial immune systems [20]. This paper focuses on genetic algorithm and ant colony optimization that are popular algorithms in evolutionary algorithms and swarm intelligence, respectively.



1. Genetic Algorithm: This algorithm was developed by John Holland and his collaborators in the 1960s and 1970s [21]. Genetic algorithm is a model or abstraction of biological evolution based on Charles Darwin's theory of natural selection [21]. Genetic algorithm involves the encoding of solutions as arrays of bits or character strings (chromosomes). The fitness of these chromosomes is evaluated. The creation of a new population is performed by the genetic operators, i.e., crossover, mutation, and fitness proportionate selection. These new chromosomes are selected based on their fitness. The old population is replaced by the new one. These steps are repeated for a number of generations. At the end, the best chromosome is decoded to obtain a solution. Figure III shows the cycle of genetic algorithm.

FIGURE III: THE CYCLE OF GENETIC ALGORITHM.

The genetic operators are the essential components of genetic algorithms as a problem-solving

strategy. Crossover operator generates 2 offspring chromosomes from parents’ chromosomes which bring about diversity characteristic of new child. There are many techniques of crossover operator operators such as single point, two point, and order crossover. Mutation operator is used to prevent falling all solutions in population into a local optimum of solved problem. Mutation changes randomly the new offspring. Selection operation is a selection of chromosomes to be the chromosome in the next generation. The best chromosome should survive and become the original breed for the next generation. Chromosomes with good fitness are probably selected. There are many methods how to select the best chromosomes, for example, Roulette wheel selection, rank selection, and tournament selection. Elitism copies the best members of the parent population to the next generation. Elitism has proved to increase speed of performance of genetic algorithm because the best found solution is preserved to the child generation [22].

2. Ant Colony Optimization: The first ant colony optimization system was introduced in 1992, and was called ant system [23]. The natural metaphor of ant algorithms are based the ant colonies [7]. Real ants are capable to find the shortest path from a food source to their nest by exploiting pheromone information. While walking, ants deposit pheromone on the ground and follow, in probability, pheromone previously deposited by other ants. These pheromone can be evaporated depend on the time. Hence, the shorter path will have more accumulated pheromone. Figure IV presents how real ants find a shortest path.



www.elvedit.com 248

In Figure IV-A, ants arrive at a decision point. Some ants choose the upper path and some ants go to the lower path. These choices are randomly selected. Since ants move at approximately constant speed, the ants which choose the lower path reach to the opposite site faster than those which choose the longer path. Pheromone accumulates at a higher rate on the shorter path. The number of dashed lines in Figure IV-D is approximately proportion of the pheromone that deposited by ants. Since the lower path is shorter than the upper one, more ants will visit it. Therefore, the pheromone accumulates faster. After a short transitory period, the difference in the total of pheromone on the two paths is sufficiently large. This pheromone is influent to the decision of the news coming ants in the system. Very soon, all ants will use the shorter path [7]. This concept can be implemented as an optimization algorithm and it can be implemented for traveling salesman problems.

FIGURE IV: PATH FINDING OF REAL ANT.

Informally, the ant colony system works by m ants. These ants are initially positioned on n

cities. The positions are chosen according to some initialization rule (e.g., randomly). Each ant builds a tour or a feasible solution to the TSP by repeatedly applying a random greedy rule or the state transition rule. While constructing a tour, an ant also modifies the amount of pheromone on the visited edges by applying the local updating rule. The amount of pheromone on edges is modified again by applying the global updating rule when all ants terminate their tour. In building the tours, ants are guided by both heuristic information and by pheromone information. An edge with a high amount of pheromone is a very desirable choice.



III. CLUSTERING EVOLUTIONARY COMPUTATION This paper proposes to use clustering techniques, to improve the performance of the

evolutionary computation algorithm, i.e., genetic algorithm and ant colony optimization, on the traveling salesman problems. This section presents the steps of solving the TSP, which can be divided into three main steps, i.e., node clustering, optimal path finding, and cluster connecting. The flow of clustering evolutionary computation algorithm is showed in Figure V.

FIGURE V: CLUSTERING EVOLUTIONARY COMPUTATION STEPS.

A. Node Clustering

In the first step, nodes in the TSP are grouped by a clustering technique. By the concept, any clustering algorithms can be used in this step. However, in this paper, K-means clustering and Gaussian mixer models will be considered because they are the distance-based clustering algorithms, which the centroids of each cluster can be easily calculated. For a TSP with 푁 nodes, the number of clusters (퐾) is defined as (2).

퐾 = 푁 2⁄ (2)

K-means clustering is one of the simplest unsupervised learning algorithms. This algorithm performs an iterative alternating fitting process to form a specific number of clusters. In the first step, a set of K points are selected to be a first guess of the means of the clusters. These points are called cluster seeds or centroids. Each node is assigned to the nearest seed to form a set of temporary clusters. Then, the seeds are replaced by the cluster means and the points are reassigned. This process continues until no further changes occur in the clusters. Hence, by the K-means clustering, the shape of cluster looks like the sphere.

In the Gaussian mixture models, the shape of cluster can be sphere or ellipse. The probability density functions of nodes are used to create the clusters. Usually a mixture model with full



www.elvedit.com 250

covariance matrices will need fewer components to model a given density, but each component has more adjustable parameters. The EM algorithm calculates the probability that each point belongs to each distribution and then uses these probabilities to compute a new estimate for the parameters. These iteration processes are continued until the parameters are not changed.

In the Figure VI, two traveling salesman problems, Berlin52 and Gil 262 [24], are visualized. The outputs of K-means clustering and Gaussian mixer model are showed in Figure VII and Figure VIII. These outputs show that there are the differences between the shapes of clusters in K-means and Gaussian mixer models when the number of clusters is predefined. The shapes of clusters in Gaussian mixer model are flexible and some clusters are ellipse. In the Gaussian mixer model, the sizes of clusters are different because they are clustered by the density of nodes, while K-means clustering will consider only the distance between node and centroid.

(a) Berlin 52 (b) Gil262

FIGURE VI: TWO TRAVELING SALESMAN PROBLEMS.

(a) K-means Clustering (b) Gaussian Mixer Model

FIGURE VII: THE CLUSTERS OF BERLIN52.



(a) K-means Clustering (b) Gaussian Mixer Model

FIGURE VIII: THE CLUSTERS OF GIL262.

B. Optimal Path Finding

After the nodes in the traveling salesman problems were clustered, the shortest path in each cluster will be searched. There are many techniques that can be used for this step. However, for the large TSP, the deterministic methods may take a lot of time. The evolutionary computation is a good choice. Genetic algorithm and ant colony optimization are the most popular evolutionary algorithms, which are widely used in many research areas. The genetic algorithm is simple and effective, while the ant colony optimization yields the good results on TSP. In this paper, the genetic algorithm is implemented according to the evolutionary theories. Permutation encoding is used to represent a chromosome that is a solution of TSP. The values in chromosomes are the order of cities which salesman will visit them. Figure IX (a) shows some examples of chromosomes in a population.

(a) Chromosome Encoding

(b) Order Crossover

(c) Inversion Mutation

FIGURE IX: CHROMOSOMES AND GENETIC OPERATORS.



www.elvedit.com 252

Only the most suitable chromosomes in a population are survived. These chromosomes are used to generate offspring, thus their biological characteristics are transmitted to new generations. The population of solutions are created by the genetic operators, i.e., order crossover and inversion mutation, that are shown in Figure IX (b) and Figure IX (c), respectively. Solutions are selected according to their fitness value to form new offspring. These processes are repeated until a fixed number of generations.

An alternative method, ant colony optimization, is implemented to solve the TSP in each cluster. Each ant constructs a TSP solution in an iterative way; it adds new cities to a partial solution by exploiting both information gained from past experience and a greedy heuristic. Ant colony optimization will be compared to the clustering genetic algorithm. Figure X shows an example of the optimal path finding in each cluster. Eleven nodes (A-K) are clustered to 3 groups and three optimal paths are founded for these three clusters. The centroid of each cluster will be used for the cluster connecting step.

FIGURE X: THE OPTIMAL PATH FINDING IN EACH CLUSTER.

C. Cluster Connecting

To connect all 퐾 paths from 퐾 clusters, a method for selecting the clusters and nodes is proposed. This method can be divided into 3 steps, i.e., the cluster selection, the next cluster selection, and the last node connection.

1. The Cluster Selection: In this step, the Euclidean distances of any two clusters are compared. All pairs of centroids will be used to calculate the distances. The shortest distance will be selected. If there are 퐾 clusters, the number of distance calculation will be (3).

푁푢푚푏푒푟표푓퐷푖푠푡푎푛푐푒퐶푎푙푐푢푙푎푡푖표푛 = !( )! !

(3)

Figure XI shows an example of the distance calculation on 3 pairs of centroids. The shortest distance is the distance between Cluster1 and Cluster2. Hence, these clusters will be used for choosing the nearest nodes.



(a) Distances of Any Two Centroids (b) The Shortest Distance

FIGURE XI: EXAMPLES OF THE DISTANCE CALCULATION AND THE SHORTEST DISTANCE. Then, the distances between Centroid1 and all nodes in Cluster2 are calculated. The nearest

node from Cluster2 will be selected, which it is node E in Figure XII (a). In the same way, the distances between Centroid2 and all nodes in Cluster1 are calculated. The nearest node from Cluster1 is selected. A path between these two nodes will give the minimum distance for two clusters connection. From the example, node A in Cluster1 will be connected to node E in Cluster2.

(a) Distances between Centroid1 and All Nodes in Cluster2 (b) Distances between Centroid2 and All Nodes in Cluster1

FIGURE XII: THE NEAREST NODES SELECTION. For node A and node E, the distance to the next left node and the distance to the next right

node will be compared. The nearest one will be selected. For example in Figure XIII (a), the distance between A and D is shorter than the distance between A and B. Therefore, node A will be connected to node D, and node D will be connected to another nodes according to the order in the optimal path from the previous step. For cluster2, node F will be connected to node E, and then node F will be connected to another nodes. Then, paths from Cluster1 and Cluster2 are connected as showed in Figure XIII (b).



www.elvedit.com 254

(a) Choosing Next Node for Each Cluster (b) A Connected Path

FIGURE XIII: EXAMPLES OF TWO PATHS CONNECTING. 2. The Next Cluster Selection: From the previous step, there are two end-nodes that are node B

and node H from Cluster1 and Cluster2, respectively. These nodes will be connected to other clusters by choosing the shortest distance between nodes and centroids. The next centroids will be selected by comparing the distances between two end-nodes and all remained centroids. This step must be repeated until the last centroid is selected as showed in Figure XIV (a).

(a) Find the Next Cluster (b) Find the Nearest Node

FIGURE XIV: THE NEXT CLUSTER SELECTION. If there are 퐾 clusters, this step will be repeated 퐾 − 2 times. The number of distance

calculation for all 퐾 − 2 repetitions will be (4).

푁푢푚푏푒푟표푓퐷푖푠푡푎푛푐푒퐶푎푙푐푢푙푎푡푖표푛 = 2(퐾 − 2) + 2(퐾 − 3) + ⋯+ 2

= 2[(퐾 − 2) + (퐾 − 3) +⋯+ 1]

= 2∑ (푖) = (퐾 − 2)(퐾 − 1) (4)



From Figure XIV, only Cluster3 have not been selected, thus the distance from node B and node H to Centroid3 will be compared. The distance between node B and Centroid3 is shorter than that of node H. Then, node B will be connected to the nearest node in Cluster3. The distances between node B and all nodes in Cluster3 are compared. Node K in Cluster3 gives the shortest distance. Therefore, node B will be connected to node K as showed in Figure XIV (b). From node K, the distances to the neighbour nodes will be compared. As showed in Figure XV (a), a shorter distance is selected, and the solution path will go along this direction.

(a) Choose the Direction (b) The Connected Path

FIGURE XV: THE NEXT CLUSTER SELECTION. 3. The Last Node Connection: At the last cluster, when there are no remaining clusters, two

end-pointed will be linked. A complete TSP solution will be appeared as showed in Figure XVI. This method may not guarantee the shortest path, because the greedy methods are applied in many steps. However, when the clustering techniques were applied, the TSP sub-problems should be easier to be solved in a specific time.

FIGURE XVI: A COMPLETE TSP SOLUTION.



www.elvedit.com 256

IV. EXPERIMENTATION In order to evaluate the proposed methods, 10 TSP datasets from TSPLIB [24] are tested.

These 10 datasets are symmetric TSPs in the two-dimensional Euclidean distance. The number of cities and the optimal tour length of these datasets are shown in Table I.

TABLE I: 10 DATASETS FOR TESTING

No. Dataset #Cities Optimal Length No. Dataset #Cities Optimal

Length

1 eil51 51 426 6 kroB200 200 29,437

2 berlin52 52 7,542 7 gil262 262 2,378

3 eil76 76 538 8 lin318 318 42,029

4 pr76 76 10,8159 9 pcb442 442 50,778

5 kroE100 100 22,068 10 rat783 783 8,806

Each experiment will be repeated ten times on each datasets to find the average minimum tour length and the average computational time. The results will be compared between simple genetic algorithm and the clustering genetic algorithms that use K-means clustering and Gaussian mixer models. Furthermore, these results will be compared to the ant colony optimization.

For the parameter settings, the number of cluster is defined as 푁 2⁄ on both K-means clustering and Gaussian mixer model, where 푁 is the number of cities in a TSP. For each cluster that is a TSP sub-problem, the parameters of genetic algorithm and ant colony optimization are defined as Table II, where 푛 is the number of nodes in a cluster.

TABLE II: THE PARAMETER SETTINGS FOR GENETIC ALGORITHM AND ANT COLONY OPTIMIZATION

Genetic Algorithm’s Parameters Value Ant Colony Optimization’s

Parameters Value

Length of Chromosomes 푛 Number of Ants 푛

Number of Chromosomes 3푛 Maximum Time 10

Number of Generations 5푛 Initial Pheromone Value 1

Crossover Rate 0.9 Influence of Pheromone on Direction (훼) 3

Mutation Rate 0.8 Influence of Adjacent Node Distance (훽) 7

Selection Method Tournament Pheromone Decreasing Factor (휌) 0.03332

Elitism Rate 0.2 Pheromone Increasing Factor (휐) 0.03332



The experimental results of simple genetic algorithm and clustering genetic algorithms are compared in Table III. The average minimum tour length and the average computational time are illustrated. The results show that most of the average minimum tour lengths of the clustering genetic algorithms, except only eil51, are better than that of the simple genetic algorithm. For eil51, the cities in this problem are randomly distributed and the number of cities is not too much. Therefore, the genetic algorithm can be used to solve that problem and yields a better result in a time limitation. For pcb442 and rat783, these problems are large and their computational times are more than 10,000 seconds. Therefore, the average minimum tour length under the pre-condition cannot be illustrated since they are time out.

TABLE III: THE RESULTS OF CLUSTERING GENETIC ALGORITHM

No. Dataset

Average Minimum Tour Length Average Computational Time (Sec.)

Genetic Algorithm

Clustering Genetic

Algorithm

Clustering

K-means GA

Gaussian GA K-means

GA Gaussian

GA

1 eil51 477 484 484 3.58 3.36 3.85

2 berlin52 8,586 8,416* 8,439* 3.67 3.81 4.35

3 eil76 641 624* 624* 12.93 3.55 4.14

4 pr76 129,456 125,243** 124,857** 12.93 3.76 4.25

5 kroE100 30,119 25,918*** 26090*** 32.56 4.04 4.46

6 kroB200 53,838 34,879*** 34,816*** 380.15 5.69 5.97

7 gil262 4,569 2,801*** 2,781*** 1,027.73 7.33 8.13

8 lin318 88,128 51,746*** 51,665*** 2,097.93 11.03 12.81

9 pcb442 - 63,851 63,970 >10,000.00 17.03 18.24

10 rat783 - 14,370 14,421 >10,000.00 41.04 39.86 * Significantly difference from the genetic algorithm at 90% confident interval. ** Significantly difference from the genetic algorithm at 99% confident interval. *** Significantly difference from the genetic algorithm at 99.9% confident interval.

Also, the results in Table III show that when the sizes of problems are increased, the results of the cluster genetic algorithms are more significantly differed from the simple genetic algorithm. On kroE100, kroB200, gil262, and lin318, the average minimum tour lengths of the clustering genetic algorithms are significantly better than those on the genetic algorithm at 99.9% confident interval. Moreover, in obviously, the average computational times of the proposed methods are shorter than those of the simple genetic algorithm when the problems are large.



www.elvedit.com 258

When the clustering techniques are compared, the results of K-means clustering is better than the results of Gaussian mixer models on 4 datasets, draw on 2 datasets, and Gaussian mixer models are better than K-means clustering on 4 datasets. For the computational time, the average times of K-means clustering is slightly better than the average times of Gaussian mixer models. Hence, when the concept of clustering genetic algorithm is used, both K-means and Gaussian mixer can be applied in the clustering step. If the data are randomly distributed, K-means clustering will be more suitable than Gaussian mixer model because it can reduce the computational time and give the better results. In the opposite, Gaussian mixer model is better than K-means when there are patterns in the distribution of data or the cities are dense in some areas of the input spaces.

When the ant colony optimization is tested, the average minimum tour lengths are lower than those of the genetic algorithm and clustering genetic algorithms. In Table IV, the results show that ant colony optimization yields the best solutions. However, the computational times of clustering ant colony optimization methods are lower for all datasets. The differences in the average minimum tour lengths are compared by the charts in Figure XVII. Since the ant colony optimization yields the good results on the average minimum tour lengths, the results may be difficult to improve by the clustering ant colony, while the clustering techniques can improve the efficiency of genetic algorithm as showed in Figure XVII (a).

TABLE IV: THE RESULTS OF CLUSTERING ANT COLONY OPTIMIZATION

No. Dataset

Average Minimum Tour Length Average Computational Time (Sec.)

Ant Colony

Clustering Ant

Colony

Clustering K-means

Ant Colony Gaussian

Ant Colony K-means Ant Colony

Gaussian Ant Colony

1 eil51 442 484 484 4.90 3.17 3.7

2 berlin52 7,608 8,424 8,515 5.25 3.57 3.95

3 eil76 559 624 623 22.35 3.36 3.97

4 pr76 115,057 125,464 125,128 20.11 3.61 4.13

5 kroE100 23,406 25,970 25,838 57.72 3.72 4.31

6 kroB200 32,276 38,960 34,538 1,044.34 5.13 5.84

7 gil262 2,647 2,752 2,747 3,113.64 7.38 7.89

8 lin318 47,514 52,093 51,641 6,992.79 11.00 11.64

9 pcb442 - 61,418 61,439 >10,000.00 17.63 18.53

10 rat783 - 13,539 13,546 >10,000.00 46.97 46.91



(a) Genetic Algorithm and Clustering GA (b) Ant Colony Optimization and Clustering Ant

FIGURE XVII: THE AVERAGE MINIMUM TOUR LENGTH. For the average computational time, both K-means and Gaussian mixer models are able to

reduce the executable times of the ant colony optimization. Hence, the average minimum tour length may be improved if the numbers of ants or the maximum times are increased. Chart in Figure XVIII illustrates that the average computational times of the simple ant colony optimization and the simple genetic algorithm are higher than those of the other methods. For ant colony optimization and genetic algorithm, they cannot be compared to each other due to the difference in their algorithms and parameters. The proposed methods, clustering evolutionary computation, can improve the average computational time of the simple algorithms. When K-means clustering or Gaussian mixer model were applied, the average computational time can be reduced, especially in the large TSP.



www.elvedit.com 260

FIGURE XVIII: THE AVERAGE COMPUTATIONAL TIME.

V. CONCLUSION Clustering evolutionary computation is proposed to improve the performance of evolution

computation on TSP. The proposed methods have three main steps, i.e., node clustering, optimal path finding, and cluster connecting. In the first step, nodes in the TSPs are divided into the smaller problems using K-means clustering or Gaussian mixer model. Then, the genetic algorithm and the ant colony optimization are used to find the optimal path. In the final step, a simple method for choosing clusters and nodes is presented to connect all clusters in the TSP. Path of all clusters are connected by considering the centroids of clusters and the marginal nodes.

The experimental results show that the clustering techniques, both K-means clustering and Gaussian mixer model, yields the better solutions when the optimal paths are found by the genetic algorithm. The computational times of the proposed methods are shorter than the basic evolutionary computation algorithm. This method is suitable for the large scale of TSP. The proposed methods can be applied to the other optimization techniques such as simulated annealing, particle swarm intelligence, or estimation of distribution algorithm (EDA). Moreover, this method is not specific for the TSPs; it can be applied to the other optimization or combinatorial problems.

REFERENCES [1] Al-Dulaimi, B.F., & Ali, H.A. (2008). Enhanced traveling salesman problem solving by genetic

algorithm technique (TSPGA). Proceedings of World Academy of Science: Engineering & Technology, 40, 296-302.

[2] Leung, K.S., Jin H.D., & Xu, Z.B. (2004). An expanding self-organizing neural network for the traveling salesman problem. Neurocomputing, 62, 267–292.



[3] Masutti, T.A.S., & de Castro, L.N. (2009). A self-organizing neural network using ideas from the immune system to solve the traveling salesman problem. Information Sciences, 179(10), 1454–1468.

[4] Lo, C.C., & Hsu, C.C. (1998). Annealing framework with learning memory. IEEE Transactions on Systems, Man, and Cybernetics, Part A: Systems and Humans, 28(5), 648-661.

[5] Jayalakshmi, G.A., & Sathiamoorthy, S. (2001). A hybrid genetic algorithm – A new approach to solve traveling salesman problem. International Journal of Computational Engineering Science, 2(2), 339–355.

[6] Tsai, H.K., Yang, J.M., Tsai, Y.F., & Kao, C.Y. (2003). Heterogeneous election genetic algorithms for traveling salesman problems. Engineering Optimization, 35(3), 297–311.

[7] Dorigo, M., & Gambardella, L.M. (1997). Ant colony system: A cooperative learning approach to the traveling salesman problem. IEEE Transactions on Evolutionary Computation, 1(1), 53-66.

[8] Liu, J.L. (2005). Rank-based ant colony optimization applied to dynamic traveling salesman problems. Engineering Optimization, 37(8), 831–847.

[9] Puris, A., Bello, R., & Herrera, F. (2010). Analysis of the efficacy of a two-stage methodology for ant colony optimization: Case of study with TSP and QAP. Expert Systems with Applications, 37(7), 5443–5453.

[10] Yang, J., Wu, C., Lee, H.P., & Liang, Y. (2008). Solving traveling salesman problems using generalized chromosome genetic algorithm. Progress in Natural Science, 18(7), 887-892.

[11] Liu, F., & Zeng, G. (2009). Study of genetic algorithm with reinforcement learning to solve the TSP. Expert Systems with Applications, 36(3), 6995–7001.

[12] Albayrak M., & Allahverdi, N. (2011). Development a new mutation operator to solve the traveling salesman problem by aid of genetic algorithms. Expert Systems with Applications, 38(3), 1313-1320.

[13] Laporte, G., & Palekar U. (2002). Some applications of the clustered travelling salesman problem. The Journal of the Operational Research Society, 53(9), 972–976.

[14] Anily, S., Bramel, J., & Hertz, A. (1999). A 5/3-approximation algorithm for the clustered traveling salesman tour and path problems. Operations Research Letters, 24(1), 29-35.

[15] Guttmann-Beck, N., Hassin, R., Khuller, S., & Raghavachari B. (1998). Approximation algorithms with bounded performance guarantees for the clustered traveling salesman problem. Lecture Notes in Computer Science, 1530, 6-17.

[16] Ding, C., Cheng Y., & He, M. (2007). Two-level genetic algorithm for clustered traveling salesman problem with application in large-scale TSPs. Tsinghua Science & Technology, 12(4), 459-465.

[17] Tan, P., Steinbach, M., & Kumar, V. (2006). Introduction to data mining. Boston: Pearson Addison Wesley.

[18] Bramer, M. (2007). Principles of data mining. London: Springer-Verlag.

[19] Dempster, A.P., Laird, N.M., & Rubin, D.B. (1977). Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society, Series B, 39(1), 1-38.



www.elvedit.com 262

[20] Das, S., Abraham, A., & Konar, A. (2009). Metaheuristic Clustering. Berlin: Springer.

[21] Goldberg, D.E. (1989). Genetic algorithms in search, optimization, and machine learning. New York: Addison Wesley.

[22] Jadaan, O.A., Rajamani, L. & Rao, C.R. (2005). Improved selection operator for GA. Journal of Theoretical and Applied Information Technology (JATIT), 2005-2008, 269-277.

[23] IRIDIA University Libre de Bruxelles. Publications of ACO algorithms [Online] 06 March 2012. Available from: URL: http://www.aco-metaheuristic.org/publications.html [Accessed 15 December 2012].

[24] TSPLIB. University of Heidelberg. Available from: http://comopt.ifi.uni-heidelberg.de/software/ TSPLIB95/ [Accessed: October 2012].


An Agent Driven M-learning Application

Authors

Collins N. Udanor Faculty of Physical Sciences, Department of Computer Science, University of Nigeria


O.U. Oparaku Faculty of Engineering, Department of Electronic Engineering, University of Nigeria.


Abstract

The future of the web is on mobile devices. Application users have migrated from the desktop to the web. Now the next stage of the Web will be building apps and mobile UIs on top of our collective data. On the part of developers, application development is moving from object-oriented development to agent-oriented programming. This paper presents a fusion of these two trends in computing. The need for ubiquitous access to information and communication, as well as the portability of devices has prompted a lot of research interests in mobile technologies. One of such recent interests is in the field of mobile learning (M-learning), an offshoot of the more established e-learning. This paper presents the development of a multi-agent driven m-learning application using the Java Agent Development Environment (JADE).

Key Words

Agent, JADE, IMLS, M-learning

I. INTRODUCTIONThe need to re-conceptualize learning in the mobile age has given rise to intensive research

work on Mobile learning. In recent times, this need has also propelled researchers to recognize the essential role of mobility and communication in the process of learning. Mobile learning is the delivery of learning contents in mobile devices over wireless infrastructures. M-learning provides learning anywhere, anytime. An agent, on the other hand, is essentially a special software component that has autonomy and provides an interoperable interface to an arbitrary system and/or behaves like a human agent, working for some clients in pursuit of its own agenda [1][12].



www.elvedit.com 264

This paper demonstrates how a mobile learning application based on the model of Intelligent Mobile Learning System (IMLS) [2] is developed.

II. REVIEW OF RELATED WORK The IMLS model is based on a modified Silander’s and Kazi’s models [3][11][4], as well as a

reference to Kinshuk’s Bee-gent framework[5]. Unlike the Bee-gent[6] multi-agent framework used by the author in [5], this work makes use of the JADE multi-agent framework, which is well researched and is gaining popularity among many researchers in agent technology. JADE is developed by Telecom Italia and offered as free software. JADE ensures consistent connection and access to M-learning contents even at low bandwidths by using bit efficient transmission technique.

Berking[7], instead of creating a new Instructional Design (ID) model, presented a framework that can be used to incorporate mobile learning considerations into existing ID models (which optimize them for the paradigm of “anywhere, anytime” mobile learning. Rather than focus on lists of specific design considerations for mobile, they created a framework that provides an organizing principle for these design considerations. Within this framework, they called out the learning theory that underlies the mobile learning strategy as an important determinant of considerations for a new or existing ID model. They stated that the flexible approach proposed by their framework takes both instruction and performance support into consideration for the mobile learning task. Botzer[8], developed a Math for Mobile application that reflected the socio-cultural and situated learning model of mobile learning. They found that the contribution of the mobile environment lies not only in making dynamic mathematical application more available, but also in supporting the execution of mathematical tasks that are closer to the students’ experiences and more relevant to them, which has the potential to enhance experiential learning. The authors [9] opined that the quality of the M-learning application represents an important aspect for the education process because it affects the way the information is understood and is learned by users. They presented M-learning application metrics. Among the utilized models of measuring the quality level of M-learning applications were the indicators such as:

Dimension of occupied space, access count of a page or topic, and number of pages read in a working session.

III. JADE AGENT BASED M-LEARNING APPLICATION DEVELOPMENT In the following sections, we shall show the design and implementation of the agent-based M-

learning application developed during our research work.

Figure 1 shows the architecture of the model of Intelligent Mobile Learning System (IMLS). The model is a multi-tier architecture, made up of five sub-units (tiers). These are: the client, the web service, the JADE agents, the IMLS Model and the ORMLite data store.

An Agent Driven M-learning Application Collins N. Udanor and O.U. Oparaku


Application Server Enterprise Servers/ Data

Browser

Request

Response

(Controller) Servlet

(View)

JSP

(Mode)

FIGURE 1: MODEL OF INTELLIGENT AGENT MOBILE LEARNING SYSTEM (IMLS)

A. The Client

This can be any mobile device such as smart phones, PDAs, tablet PCs, etc. For the prototype application based on this model, the client is built using the Android smart phone. The mobile device connects to the Internet using a WAP browser on basic HTML, XML, Java scripts and Cascaded Styling Sheet (CSS) for presentation. The Internet provides connection to the server end of the model. XML helps to communicate with the agents in the server to collect data and return the results to the calling session. JavaServer Pages (JSP) technology handles the business logic. JSP is well on its way to becoming the preeminent Java technology for building applications that serve dynamic Web content [10].

It is a server-side scripting language. The biggest advantage of using JSP is that it helps effectively separate presentation from content. Architecturally speaking, JSP can be viewed as a high-level abstraction of servlets that is implemented as an extension of the Servlet 2.1 API.

FIGURE 2: JSP MODEL 2 ARCHITECTURE



www.elvedit.com 266

The early JSP specifications advocated two philosophical approaches for building applications using JSP technology. These approaches, termed the JSP Model 1 and Model 2 architectures, differ essentially in the location at which the bulk of the request processing was performed. In the Model 1 architecture, the JSP page alone is responsible for processing the incoming request and replying back to the client. There is still separation of presentation from content, because all data access is performed using beans.

In the IMLS model, the Model 2 architecture, shown in figure 2, is employed. The model 2 is a hybrid approach for serving dynamic content, since it combines the use of both servlets and JSP. It takes advantage of the predominant strengths of both technologies in using JSP to generate the presentation layer and servlets to perform process-intensive tasks. There is still separation of presentation from content, because all data access is performed using beans. Here, the servlet acts as the controller and is in charge of the request processing and the creation of any beans or objects used by the JSP, as well as deciding, depending on the user's actions, which JSP page to forward the request to. Note particularly that there is no processing logic within the JSP page itself; it is simply responsible for retrieving any objects or beans that may have been previously created by the servlet, and extracting the dynamic content from that servlet for insertion within static templates.

B. The Server, Web Services Producer

This is a Web server, typically behind firewalls and/or proxy gateways. The Web Services Integration Gateway (WSIG), an add-on component to the JADE platform provides an interface between the JSP client and the agent world [12]. The objective of WSIG is to expose services provided by agents and published in the JADE Directory Facilitator (DF) as web services with no or minimal additional effort, though giving developers enough flexibility to meet specific requirements they may have. The process involves the generation of a suitable Web Service Description Language (WSDL) for each service-description registered with the DF and possibly the publication of the exposed services in a UDDI registry. The WSIG add-on supports the standard Web services stack, consisting of WSDL for service descriptions, Simple Object Access Protocol (SOAP) message transport and a UDDI repository for publishing Web services. The WSIG web application consists of two main elements:

• WSIG Servlet

• WSIG Agent

The WSIG Servlet is the front-end towards the internet world and is responsible for:

Serving incoming HTTP/SOAP requests

Extracting the SOAP message

Preparing the corresponding agent action and passing it to the WSIG Agent.



Converting the action result into a SOAP message

Preparing the HTTP/SOAP response to be sent back to the client

The WSIG Agent is the gateway between the Web and the Agent worlds and is responsible for:

Forwarding agent actions received from the WSIG Servlet to the agents. It is actually able to serve them and get back responses.

Subscribing to the JADE DF to receive notifications about agent registrations and de-registrations.

Creating the WSDL corresponding to each agent service registered with the DF and publishes the service in a UDDI registry if needed.

Two main processes are continuously active in the WSIG web application:

The process responsible for intercepting DF registrations/deregistrations and converting them into suitable WSDLs. As mentioned, this process is completely carried out by the WSIG Agent.

The process responsible for serving incoming web service requests and triggering the corresponding agent actions. This process is carried out jointly by the WSIG Servlet (performing the necessary translations) and the WSIG Agent (forwarding requests to agents able to serve them).

JADE agents publish their services in the DF providing a structure called DF-Agent-Description as defined by the FIPA specification (www.fipa.org) [13]. A DF Agent-Description includes one or more Service-Descriptions, each one actually describing a service provided by the registering agent. A Service-Description typically specifies, among others, one or more ontologies that must be known in order to access the published service.

The actions the registering agent is actually able to perform are those defined in the specified ontologies.

C. The IMLS Model sub-module

The model consists of the components that make up the Intelligent Tutoring/Learning Systems. These include the student model, expert module, pedagogical module and the communication or user interface module as depicted in figure 3. The model contains the actual course content, the student records, assessment, as well as administrative information. These contents are stored in the data store using the ORMLite. To interact with these contents, the respective agents are invoked, such as the authentication agent, tutorial agent, assessment and user agents.



www.elvedit.com 268

IV.

FIGURE 3: IMLS MODEL SUB-MODULE

D. The JADE Agent

The system relies heavily on multi-agents, which ensuring effective bandwidth utilization optimizes the processing time. Agents are software components that have been widely used in the fields of artificial intelligence, database management, computer networking, etc. These agents have autonomy, provide an interoperable interface to an arbitrary system and behave like human agents, working for some clients in pursuit of its own agenda. The IMLS model in figure 1 also contains the JADE framework which consists of the Agent Management System (AMS) and DF, in addition to the IMLS agents created by the programmer. These agents include the Authentication agent, User agent, Tutorial agent and Assessment agent. The Authentication Agent manages user accounts, privileges and sessions. Before the server starts running, the JADE platform is first booted with the AMS and the DF on port 1099 as shown on the left panel in figure 4.

The user agent keeps track of each user’s activities after the user has been authenticated. The agent stores session information, tutorial sessions and assessment information. Figure 5 shows a user login and a user session on a web browser. A sniffer agent in the JADE platform keeps track of the interactions between the various agents, as shown on the right panel of figure 4. The tutorial agent delivers the course materials to the user, while the assessment agent delivers test questions to determine the learner’s performance. The assessment could be a pre-lesson or post lesson assessment.

IMLS Model sub-module

Student

Model

Student

Pedagogical

Module

Expert

Module

Communication Module (U.I)

User Agent

Tutorial Agent

Administrator

Assessment

Authentication



FIGURE 4: JADE PLATFORM RUNNING THE AUTHENTICATION AGENT AND WSIG SERVLET.

FIGURE 5: A USER SESSION

FIGURE 6: ADMINISTRATOR BACKEND



www.elvedit.com 270

E. The Data Store

This unit is made of the Object Relational Mapping, and a relational data base management system such as MySql, MS-SQL Server, Oracle, etc. This application uses a MySql database.

ORMLite 4.3.1

ORMLite provides a lightweight Object Relational Mapping between Java classes and SQL databases [14]. There are certainly more mature ORMs which provide this functionality including Hibernate and iBatis. ORMLite supports JDBC connections to MySQL, Postgres, H2, SQLite, Derby, HSQLDB, Microsoft SQL Server, and can be extended to additional ones relatively easily. ORMLite also supports native database calls on Android OS. There are also initial implementations for DB2, Oracle, generic ODBC, and Netezza.

F. The Network

This refers to the wireless and wired networks, part of the Internet, and the communication protocols. The network (Internet) connects the mobile device (web service consumer) to the Web services producer via SOAP/XML. It is important to note that JSR 172 [15] does not mandate the use of XML encoding on the device itself, allowing implementations (as long as they are transparent to both consumer and producer) to use more efficient encoding approaches, such as the use of binary protocols between the device and the wireless gateway.

IV. CONCLUSION Agent-oriented application development promises the delivery of software models that consist

of dynamically interacting rule-based agents. The systems they interact in can create real-world-like agents that are complex, intelligent and purposeful. This system simulates the real world classroom scenario with the tutorial agent acting as a class teacher. Agents have been found to be very useful in complex systems such as data mining, ecological sciences, life and social sciences.

REFERENCES [1] Fabio Bellifemine, Giovanni Caire, Dominic Greenwood (2007). Developing Multi-Agent Systems with

JADE. John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England.

[2] Collins N. Udanor (2011). An Agent-based Model of Intelligent M-learning System. International Journal of Science and Advanced Technology. Volume 1 Number 5. Pages 65-73.

[3] Pasi Silander , Anni Rytkönen (2007) An Intelligent Mobile Tutoring Tool Enabling Individualisation of Students’ Learning Processes. Häme Polytechnic University of Applied Sciences and University of Joensuu, Finland ([email protected]) Department of Computer Science, University of Helsinki, Finland ([email protected])

[4] Sabbir Ahmed Kazi (2005). VocaTest: An Intelligent Tutoring System for Vocabulary Learning using the "mLearning" Approach. Centre for Research in Pedagogy and Practice National Institute of



Education. Available on: http://conference.nie.edu.sg/paper/Converted%20Pdf/ab00283.pdf. Viewed on 20th March, 2010.

[5] Kinshuk, Taiyu Lin (2004). Improving mobile learning environments by applying mobile agent technology. Massey University, Palmerstone North, New Zealand. Available on: http://www.col.org/pcf3/papers/pdfs/kinshuk_lin_2.pdf. viewed 12th March, 2010.

[6] Toshiba (2005). Multi-Agent Framework for 100% Pure Agent System .Corporate Research & Development Center TOSHIBA Corporation. Available on: http://www.toshiba.co.jp/rdc/beegent/index.htm. Viewed on 21st April. 2010.

[7] Peter Berking, Jason Haag, Thomas Archibald, Marcus Birtwhistle (2012). Mobile Learning: Not Just Another Delivery Method. Interservice/Industry Training, Simulation, and Education Conference (I/ITSEC) 2012

[8] Galit Botzer, Michal Yerushalmy2007). Mobile Application for Mobile Learning. (IADIS International Conference on Cognition and Exploratory Learning in Digital Age (CELDA 2007)

[9] Catalin Boja, Lorena Batagan (2009). Software Characteristics of M-Learning Applications. WSEAS Transactions on Computers, Volume 8, Issue 5, Pages 767-777

[10] Govind Seshadri. Understanding JavaServer Pages Model 2 architecture exploring the MVC design pattern. See:http://www. JavaWorld.com, viewed on 18th June, 2012.

[11] Kassim, A. A., Kazi, S. A. and Ranganath, S (2004)., “A Webbased Intelligent Learning Environment for Digital Systems”. International Journal for Engineering Education, Vol 20, No 1, pp 13-23.

[12] Stuart J. Russell and Peter Norvig (2003). Artificial Intelligence: A Modern Approach. Second Edition, Pearson Education, Inc. Upper Saddle River, New Jersey 07458.

[10] Bellifemine, F., Poggi, A. and Rimassa, G. Developing Multi Agent Systems with a FIPA-Compliant Agent Framework. In Software–Practice and Experience, John Wiley & Sons, Ltd, vol. 31, pp. 103–128,

[13] [FIPA00001] FIPA Abstract Architecture Specification. Foundation for Intelligent Physical Agents, http://www.fipa.org/specs/fipa00001/

[14] Gray Watson. OrmLite -- Simple Object Relational Mapping (ORM) Java Package . Available at: http://256.com/gray/ simple_orm_java.shtml. Viewed on 16th June, 2012.

[15] C. Enrique Ortiz, Web Services APIs for J2ME, Part 1: Remote service invocation API. Available on: www.J2MEDeveloper.com. Viewed on 5th January,2011.


Dr. Collins Udanor has his B.Eng in Computer Science and Engineering, 1996, an M.Sc in Computer Science (Data Communication option), 2004 and a PhD in Electronic Engineering, University of Nigeria in 2013 (Digital Computer Systems option, with specialization in Multi-agents based Mobile Learning). He has worked over a decade as both a Technical Instructor and as an Academic. His research interests are in the areas of intelligent mobile learning systems, Distributed Computing Infrastructures and High

Performance Computing (DCI/HPC). Over the years he has been involved in teaching, research, workshop



www.elvedit.com 272

organization and presentations. He has also been richly involved in software development projects of various capacities and on varying platforms. He has also designed and implemented Distributed Computing Infrastructures (DCIs), notably the LIONGRID-UNN-NG.

Prof. Ogbonna Ukachukwu Oparaku is a 1980 graduate of Electrical/Electronic Engineering from the University of Nigeria. He worked with General Electric Company (Telecomminications Division) Nigeria Limited between 1981 and 1983 as a Special Projects Engineer before joining the Universty of Nigeria, on parallel appointment in the Electronic Engineering Department and the Energy Research Centre. He obtained a PhD in Solid State Electronics from the University of Northumbria at Newcastle –Upon-Tyne, United Kingdom, in 1988. He became a Professor in 2003 and was

appointed Director of the National Centre for Energy Research and Development in 2004, a position he occupied till October 2009. He is currently the Head of Department of Electronic Engineering, University of Nigeria, Nsukka.


Evolution of Utilizing Multiple Similarity Criteria in Web Service Discovery

Authors

Hassan Azizi Darounkolaei Department of Computer Engineering/ Islamic Azad University, Babol Branch, Iran

[email protected] Babol, Iran

Seyed Yaser Bozorgi Rad Department of Computer Engineering/ Islamic Azad University, Babol Branch, Iran


Abstract

With the increasing use of web service in distributed platforms such as the Internet, the importance of the discovery process improvements has been added. Hybrid matchmaker methods and semantic web service, have tried to improve it further. Efficiency of matchmakers that have used text similarity criteria for the purpose has largely depends on the selection criteria. Given the importance of this topic, main idea is simultaneous use of several similarity criteria in process of web service discovery that for this purpose, a method of using multiple similarity criteria for calculating the similarity between the input/output parameters of the web service, offered. Thus, improvements resulting from the performance of any of various similarity criteria can be aggregated and better overall results for the entire set of queries, obtained. Also, according to some characteristics of web services, two asymmetric similarity criteria, introduced. And a new method for aggregating similarity of input and output parameters of web services, provided. The result of applying the proposed method shows the best performance in general, compared with the result of applying the similarity criteria separately. Also compared with two matchmakers raised in this context, the proposed method has shown better performance.

Key Words

Web Service Discovery, Hybrid Matchmaking Web Service, Semantic Web Service, Text Similarity Criteria, Asymmetric Similarity Criteria, Ordered Weighted Averaging.



www.elvedit.com 274

I. INTRODUCTION Internet since the beginning if its career, has undergone many changes, some of which in recent decades has changed human lifestyle. One of the latest changes in internet usage, with the introduction of web service has been done. Some experts predict that its impact on the Internet is high enough. So believe that the Internet will become to a network of services from a network of static documents. One glance at the web service concept, we find that this statement is not too surprising, and indeed, this concept can have a profound impact on the Internet and subsequently on various aspects of life.

As noted, one of the major challenges in the field of web services is to find the requester service through a lot of description of web service offered by different providers [1]. To describe web services, there are several ways [2] that is effective in service discovery method [3]. Given the importance of determining the measure of similarity between web services to use in discovery process, different methods have been proposed. But a measure that could make all the different scenarios to result the best solution is not accepted [4]. So it seems the best way is to use a method for applying multiple similarity criteria to achieve better result in the service discovery process. Naturally, any change and new concept in the world of technology has its own problems and complexities. Taking advantage of web services is no exception to this role and experts of this field has forced many challenges that including such items as web service discovery, automatic composition and automate monitoring application process.

II. SERVICE AND WEB SERVICE Before anything else, should be familiar with the concept of service. In the discussed field of this paper, services associated with the concept of service-oriented computing, And this concept represent a new generation of distributed computing platform that includes many items such as templates and specific design principles and also distinctive architecture model and concept, technologies and frameworks associated with it. Service orientation is a design pattern to create logical and separate units of resolve. So that the units can be used frequently and they can achieved specific objectives with contribution. And therefore, with these descriptions, a service is a unit of logic in problem solving, where the concept of service orientation has been achieved at an acceptable level [5]. After understanding service concept in the discussed field, web service will be defined. According to the definition provided for the service, the web service introduced simply, web service is implementing a service that is reusable and invoked through the web. It is a long time that client/server architecture is used for distributed applications as a web service has a similar structure. A web application is to meet a specific need in the supplier (server) that makes it possible to call and execute through the web for another program in the requester (client) [6]. On a more practical definition by W3C, web service is introduced as follows [8]:

“A web service is a software system that is design to support communication between machines on a network, for their cooperation, and includes an interface that is described in a machine-readable format (WSDL)”.

Evolution of Utilizing Multiple Similarity Criteria in Web Service Discovery Hassan Azizi Darounkolaei and Seyed Yaser Bozorgi Rad


Other systems interact with the web service in a manner which is prescribed in the relevant description and it is done by using SOAP messages that typically are sent over Http serialize Xml and other web related standards. For years web services architecture as a stack of Xml-based standards such as WSDL,SOAP and UDDI was proposed and accepted by the business and academic communities, but in recent years there is a new look based on Fielding’s doctoral thesis [8] - that it introduced a REST architecture pattern- and it’s trying to reduce the complexity of previous look. Accordingly, two major classes of web services could be understood [7].

Web services based on RESTful in which the main goal of the service is handling the display of Xml of web resources by a uniform set of stateless acts. Large or optional web services where the various standards used. And a service may offer a discretionary. There are many ways to integrate the major functional systems. As noted, in the case of web service, nowadays, two main methods have been proposed. In a way, a big stack of related technologies (SOAP, WSDL, etc.) makes it possible. But another solution based on doctoral thesis of Fielding [8] presented the implementing RPC on web that RESTful web service name is known. And not only because of their use in APIs of many web 2 service, but for the ease considered in the dissemination and use of these types of web services, is attracting attention [9]. In the field of resource-oriented architecture (ROA) and service-oriented (SOA), distributed components, respectively, are called resources and services. In the next section, these two classes of web services introduced briefly. After the emergence of RESTful web services, many names have been used to classify types of web services. And in various resources for classes that are in the RESTful web service, titles such as optional web service –proposed in the W3C categories [8]- large web services or WS-*, are used. However, the obvious characteristic of this class of web services is using a stack of related technologies and complexity compared to other methods. The main goal of the service-oriented architecture and web services technology is creating seamless interoperability between heterogeneous firmware technology stack and developing loose coupling18 in consumer service (customer, requester) and supplier of the service. And according to the modularity and the ability to combine these methods, it have been lead to a large set of WS-* specifications.

First, REST is introduced as an architectural style for creating large distributed systems. And it was used for extensibility of the Http protocol. So in many cases, the term REST, is considered in relevance with Http web services that are called RESTful, in fact, follow the REST principles and they are built based on them that simplicity, one of the most characteristic of them. These web services in close contact with the source and source-oriented architecture. Resource-oriented architecture is built based on the resource. The source is a distributed component that can be accessed directly and is managed by a common standard interface. RESTful web services that based on the development of REST technology make it possible to create resource-oriented architecture [10]. In this architecture, resources are fundamental objects. RESTful web services seem simple because of the known and existing benefit and therefore, the required infrastructure is already available pervasively. For all major programming languages, Http client and server, as well as OS\hardware platforms, is available.



www.elvedit.com 276

And Http port 80 in the firewall settings, largely by default, is left open. Such an infrastructure made possible to easy construct a client to connect to the REStful service. And the conventional browser can be used to test it, without requirement of client-side special programming. REST has proven that it can also be detected by using URL19s and hyperlinks, without the need to register a service in a particular store [9].

III. STUDY OF WEB SERVICE DISCOVERY METHODS Web service discovery methods can be classified in to different aspects. According to the subject of this research, study of detection methods in terms of their use of both techniques semantic web and information retrieval is very important. So, in this section, the various methods that have used these techniques, studied. In general, these methods can be divided in to two categories: methods based on logic and methods that are not based on logic. Some division based on the type of web service description such as [11], but it should be done based on the techniques used. In some works such as [12], techniques of the information retrieval area have been applied on semantic web service. In some works such as [13], hybrid methods that are used in several ways, introduced. In matchmaking based on the logic of web service, semantic description must necessarily be used. In most categories, these methods are among the most semantic methods. According to the ontology used to describe web services, different methods are used for this purpose, but usually some form of logical inference are used. Most methods of semantic web services matchmaking act inductively20 that the matchmaking method is based on logic. And in this case, the main idea that is to detect semantic relationship between the sources is maintained. And this matter is achieved by the service input and output, preconditions and post-conditions (or the effects) in the web service description. And these cases are part of the functional properties of service. Some non-functional characteristics of web services, such as quality of service, can be described by a service profile that some matchmakers benefit these non-functional characteristics to present final result. The other method is process-oriented matchmaking21 that compares two services in terms of their operational behavior of control flow and data flow aspects. But this method is rarely used because of lack the full support of the current format of semantic web service description. Most of the methods that benefit service profiles compare pairs of descriptions based on the logic of service profile meanings. For this, the concept of logic and rules taken from relevant ontology and service provider or service requester of different ontology, must somehow be integrated at design time or run time.

IV. THE INTRODUCTION OF THE PROPOSED METHOD Matchmaking logical methods of semantic web service alone could not meet needs. And the use of text similarity criteria considerably improves the process. Due to this, the main idea is based on using a textual similarity criterion simultaneously in web service matchmaking in order to increase the enjoyment positive impact of these criteria. In the proposed method, the similarity between service requests and service offers, are measured by using different text similarity criteria. In the proposed method, instead of using a similarity criterion, several measures are used simultaneously that by considering the differences caused by using the different criteria, in



the final result presentation, aggregation of all used criteria be benefited. Differences resulting from different criteria can be verified from the following two aspects:

1- The used model to represent the input/output parameters.

2-The method for calculate similarity.

Although the text similarity criteria used, however, a description language for the semantic web service used. The main reason is the lack of enough words in the non-semantic described language for web service to apply these methods. And another major reason is considerable increase in the accuracy of these methods when using semantic descriptions [13], [14], Because in this case the used terms that offered to describe a web service and request using ontology developed. An example used as shown in chapter 3. But it will not reduce the generality of this approach. And this method will also be used on non-semantic web services. In this work, used matchmaking framework that is used in [15] where possibility of benefiting from textual similarity criteria have been considered. In this context, there is the possibility of defining and using the valuator of different matchmaking terms. The following section will introduce used matchmaking terms. This framework consists of three main parts:

Service information extractor 22

Matchmaking expression valuator 23

Ranking processor 24

The section of the service information extractor has capabilities of processing the semantic web services to the OWL-S language.

The task of the matchmaking terms valuator is assessing and matchmaking the description of offered services with application service in accordance with the procedure specified in the matchmaking statement. The matchmaking action takes place in this section. And in this work, the proposed method has been implemented in this field. Finally, ranking processors is responsible for the final ranking of the provided services based on output of matchmaking expression valuator section. The task of service information extractor is extracting data from the offered web services description and the requested service. In this section, the API that has been implemented in version 1/1 based on the ontology of the OWL-S, used. So evaluated web service description should be based on OWL-S. Requested service should be logged with OWL-S semantic web services description language. Service request is like providing a description of a new semantic web service in OWL-S language to the system, which is regarded as the requested service. Information extraction based on the selected parameters for calculating similarities such as input parameters, output parameters, preconditions, effects, description and quality of service, performed. In this work, only two input parameters and output parameters are used. Because, these two parameters are considered functionality as important features and yet a lot of semantic



www.elvedit.com 278

web service lack features such as pre-conditions in their descriptions [12]. Information extraction must be done by using a common ontology among the requester, provider, and matchmaker. In the framework used in this work, this ontology was created and developed each time by adding semantic service descriptions to a provided service or requested. Finally, the information extraction is performed by detecting concepts in the input and output parameters, as shown in chapter 3, by using a common ontology, more information can be presented to [15]. Main stage of matchmaking act and discovering web service in the matchmaking expression valuator section is done. This is done by matchmaking term definition to apply on the extracted information from the web service descriptions in the previous step. How-to define these terms, specifies the performance of web service matchmaker. In the proposed method, several similarity criteria and a way for aggregation as the matchmaking term is used. In the final stage, the result of the matchmaking term valuator section that includes of the level of the final similarity between requested web services and individual descriptions of provided web service available in the store, are sorted by the ranking processor and ordered list as the final output of the system is presented. By revealing the influence of using the textual similarity criteria in the web services matchmaking, in this work, a method is presented which allows simultaneous use of several different similarity criteria. Overall, the performance of this method is that, first, by using two different models, representation of the parameters involved in calculating the similarity between descriptions of two models, similarity between requests and available services, calculated. Then the results of these methods aggregated. It should be noted, in this work only the input and output parameters of web service are used to calculate similarities. Furthermore, in the proposed method, a new function for aggregating the similarities of these two parameters is introduced. In this work, also two new asymmetric similarity criteria for web services are presented. Two factors are involved to make a difference in the results of the use f text similarity criteria. One of these factors is the used model to represent the parameters of the extracted information from the descriptions of the web service. In this work, two models are used for this that how-to extract them are introduced and explained.

V. RESULT OF THE SURVEY This study examines the differences resulting from the application of different similarity criteria in the process of discovery and matchmaking web services. As noted, the similarity measuring of the process of discovery and web services matchmaking in very important to have a great impact on the result of matchmaking. In this paper, a method for simultaneous use of several similarity criteria to measure the similarity between the web services offered. According to the survey that was conducted in this study, using a similarity criterion alone could not cause all queries to obtain best results. And each of the similarity criteria in some of the queries does better. Given those findings, using the same method that uses several similarity criteria can partly solve this defect. Therefore, after considering the characteristics of web service discovery problem, in this paper, a multimode decision method that is called ordered weighted averaging was used to take advantage of multiple similarity criteria. Using this method, in addition to being consistent with the characteristics of web service discovery issue as mentioned in chapter



4, different options in terms of decision optimism are provided for the user that by using the quantifiers of the natural language, to determine the weight vector in this way, these options will be based on natural language. That is cause to ease of operation for the user to select.

The proposed method is evaluated by various criteria. It was found that this method is able to create a balance between the results of different similarity criteria. And cause the difference between the best and worst performing queries in the general case is reduced by considering all the queries in the test set. And even in some of the queries can act better than all the used similarity criteria. The proposed method in this work has high expandability. And in this method, a variety of similarity criteria can be used easily. And their numbers increase or decrease. The user can easily by selecting fuzzy quantifier of natural language, reduce the degree of method optimism or increase. This will affect the matchmaking result.

In this study, in addition to the above methods, the introduction of two new asymmetric similarity criteria to calculate the similarity degree of input and output parameters of web services is mentioned. These two criteria are introduced by evaluating the specific characteristics of web service and the state of mind of some rules in this regard. After evaluating these standards, optimal performance in comparison with the standard symmetric similarity criteria that used in information retrieval will be shown. Besides, the final integrated approach of similarity between the input and output parameters of a web service were considered in this study. For this purpose, by evaluating the specific characteristics of web services, a new function for aggregating similarity value of input and output parameters, introduced. In this function, by considering the importance of two parameters of input and output, different weigh for the similarity of these two parameters is assigned. And the degree of the similarity is calculated. Bye comparing this method with simple averaging method that has been used in many other things, it was found that this function shows the better relative performance.

The main innovation of this research is the simultaneous use of several textual similarity criteria for measuring the similarity of input and output parameters of web services. Due to the different functions of each criterion against various queries, this method has been used. So all used similarity criteria influence the final result. Another innovation in this study is the introduction of two new asymmetric similarity criteria to measure the similarity of input/output parameters of web 2 service that is presented based on some characteristics of web services. The third innovation of this research is related to the way of final aggregating of the similarities of input and output parameters of a web service. Instead of simple averaging, assigns a weight to similarity level of each parameters based on a new function. The manner of the weight determination is based on some mental rules that are offered in the case of web services.



www.elvedit.com 280

VI. CONCLUSION This paper performed experiments and expressed reports and their related analyzes to evaluate hypothesis examined in this research. The obtained result point out that none of the used textual similarity measures absolutely could not alone cause all queries to achieve better results than other measures. None of similarity criteria could not introduced as the best measures. It was also shown that the proposed method in the general case for all queries can cause partial outcomes better than using a similarity criterion. This means that it reduces the difference between the results of best and worst performance of different similarity criteria, for each of the queries. Comparison of proposed method with two others matchmaking has shown that this method acts better than the others. In addition, the result of conducted experiments for evaluating the performance of two new asymmetric criteria of similarity, which is introduced in this work, was presented. The results indicate good performance of these criteria in measuring the level of the similarity between the input and output parameters of a web service that causes significant improvements of performance matchmaker in some queries by using this similarity criteria than other symmetric similarity criteria.

References

[1] Junghans M, Agarwal S, Studer R. (2010). "Towards practical semantic web service discovery". In: Proceedings of The Semantic Web: Research and Applications, 7th Extended Semantic Web Conference, ESWC 2010, Heraklion, Greece, pp. 15-29.

[2] Torma S, Villstedt J, Lehtinen V, Oliver I, Luukkala V. (2008). "Semantic Web Services — A Survey".

Helsinki: Helsinki University of Technology, Laboratory of Software Technology. [3] Liu W, Wong W. (2009). "Web service clustering using text mining techniques". International Journal

of Agent-Oriented Software Engineering, Vol. 3, No. 1, pp. 6-26. [4] Skoutas D, Sacharidis D, Simitsis A, Sellis T. (2010). "Ranking and clustering web services using

multicriteria dominance relationships". IEEE Transactions on Services Computing,Vol. 3, No. 3, pp. 163-177.

[5] Erl T, Karmarkar A, Walmsley P, Haas H, Yalcinalp U, Liu C. (2008). "Web Service Contract Design

and Versioning for SOA". Boston: Prentice Hall. [6] Yu L. (2007). "Introduction to the Semantic Web and Semantic Web Services". Boca Raton, Florida:

Chapman and Hall/CRC. [7] W3C. (2004). "Web Services Architecture". Retrieved 2011-12-22, from http://www.w3.org/TR/ws-

arch/ [8] Fielding R. (2000). "Architectural Styles and the Design of Network-based Software Architectures".

Doctor of Philosophy, University of California, Irvine. [9] Pautasso C, Zimmermann O, Leymann F. (2008). "RESTful Web Services vs. Big Web Services:

Making the Right Architectural Decision". In: Proceeding of the 17th international conference on World Wide Web (2008), Beijing, China, pp. 805-814.



[10] Garofalakis J, Panagis Y, Sakkopoulos E, Tsakalidis A. (2006). "Contemporary web service discovery mechanisms". Journal of Web Engineering, Vol. 5, No. 3, pp. 265-290.

[11] Klusch M, Fries B, Sycara K. (2009). "OWLS-MX: A hybrid Semantic Web service matchmaker for

OWL-S services". Web Semantics: Science, Services and Agents on the World Wide Web, Vol. 7, No. 2, pp. 121-133.

[12] Klusch M, Kapahnke P. (2012). "Adaptive signature-based semantic selection of services with OWLS-

MX3". Multiagent and Grid Systems,Vol. 8, No. 1, pp. 69-82. [13] Nayak R, Lee B. (2007). "Web service discovery with additional semantics and clustering". In:

Proceedings of the IEEE/WIC/ACM International Conference on Web Intelligence, WI 2007, Silicon Valley, USA, pp. 555-558.

[14] Kapahnke P, Klusch M. (2012). "Adaptive Hybrid Selection of Semantic Services: The iSeM

Matchmaker". In: Semantic Web Services, pp. 63-82: Springer Berlin Heidelberg.


Multi-Aspect Tasks in Software Education:

a Case of a Recursive Parser

Author

Evgeny Pyshkin Institute of Computing and Control / St. Petersburg State Polytechnic University

[email protected] St. Petersburg, 195251, Russia

Abstract

In this paper the task of parsing arithmetic parenthesis-free expressions parsing is investigated with special emphasis on their using in software education. As a kind of authentic problem from the areas of text processing and data structures, the task of recursive descent parser construction illustrates basic concepts of syntactic analysis and code execution. It is complex enough for explanation of parsing methods; it may be considered as an example of software characterized by the complexity of its logical structure. At the same time this task is still manageable to meet the academic requirements. In this paper we show how to introduce a synterm concept and describe an approach to lexer construction. We highlight parser source code constructions implementing grammar rules in a way that programming control structures match syntax diagram structures. Graphic formalisms are used to represent the source code structure independently of both an implementation language and a software paradigm. A variant of an object-oriented solution can serve as an example of relatively complicated entity relationship structure what makes it suitable for classroom discussion.

Key Words

Computer Science Education, Software, Syntactic Analysis, Software Structure, Task-Oriented Study.

I. INTRODUCTIONAs far as 1980 Shneiderman introduced basic types of software complexity which include

complexity of logical, structural and psychological nature [1]. Logical complexity relates to numerous conditional statements and unobvious relationships between these statements. One of the logical complexity consequences is that it is impossible or at least unevident to prove software program correctness in most practical cases. Using visual models and representations is one of

Multi-Aspect Tasks in Software Education: a Case of a Recursive Parser Evgeny Pyshkin


known ways to decrease software logical complexity.

Being one of typical processes in software design and compiler theory, text processing remains an important part of various data manipulations. Text processing algorithms encompass a considerable domain in information systems whether it concerns formatted texts based on structured representations (e.g. XML) or less structured narrative texts. In computer science education the task of building text parsers is used not only while teaching translating systems but also while teaching programming and software engineering courses. The problem of parsing an arithmetic or logic expression is far from having even a gloss of novelty [2]. Parsing expressions defined by a recursive context-free grammar is a kind of authentic problem illustrating basic concepts of lexical and syntactic analysis and code execution and to discover and test lexical and syntactic algorithms. The task is complex enough to examine parsing methods without risk to lose significant details related to types of software with essential logical complexity. At the same time it is suitable to correspond with the academic needs. That’s why the task of arithmetic expression processing is often used to expain language grammars [3], recursive data structures [4] as well as algorithms of automated syntactic analysis [5]. It may even serve as a simple model of a computational system [6] which allows meeting most intrinsic problems that developers face when they work with more complex languages. Another area where an importance of factored functions processing couldn’t be overestimated is multi-level logic synthesis, where factored expressions are usually obtained as an output of minimization and factorization procedures [7].

Constructing an expression parser can be used as an example of relatively complex component of a framework where precise and efficient formal parsers are combined with partially structured natural language parsers [8]. Thus, the task of constructing an arithmetic expression parser and expression executor is definitely multi-aspect, and its solution requires using a variety of fundamental concepts of informatics and programming.

In every programming course students learn the concept of syntax-oriented translation [9]. It is commonly known that text processing usually includes two relatively independent procedures: lexical analysis (tokenizing) and syntactic analysis (parsing). Decomposing the process to these stages significantly decreases the complexity of program code translation. In recent decades understanding of lexical, syntactic analysis and code generation as consequent or collaborating procedures is shifted to the component centered model (and further – to the service oriented model). Based on their functions compiler components provide interfaces to an external developer framework and give tools for such services as keyword highlighting, source code line partial parsing, debugging, etc. Nevertheless, from the point of view of the educational methodology, learning lexer and parser components as consequent and cooperating procedures facilitates adopting the study material to the academic needs.

The task of expression parsing (as a simplified compiler-like problem) may be used to illustrate not only programming techniques and data structures, but software engineering principles as well. Following Sommerville’s definition from [10] there are the following software engineering principles (listed here with brief description of their relation to the area of compiler construction and learning):

Strictness and formality. Correctness is critical. Solutions are based on theories:



compiling, formal grammars, languages, automata theory, data typing. Task decomposition. The analysis include own compiler components, user interface

components, profiling tools. Modularity. The main models to be investigated are the following: lexical analyzer,

parser, code generation component, etc. Abstraction. There is a space to introduce concepts of abstract syntax as well as abstract

computation models and virtual machines. Changeability. In compiler development changeability is critical due to the hardware

and language standard changes. Generality. The intermediate code for platform independence can serve as an example of

how compiler construction corresponds to the generality principle. Incrementality. Incrementality deals with aspects of versioning, code optimization and

library changes.

We consider this paper to be a continuation of the discourse on programming teaching presented in [11] with paying attention to more complex examples of code construction which refer not only to classroom situations but to the engineering practices as well.

II. THE TASK: DEFINING AN EXPRESSION GRAMMAR The syntax of a parenthesis-free expression may be defined by a context-free grammar 퐸 =

(푉 ,푉 ,푃, 푆) where 푉 is a finite set of nonterminal symbols, 푉 is a finite set of terminal symbols, 푆 is a start symbol, and 푃 is a finite set of the grammar production rules. The first group of rules is to be used during the syntactic analysis stage:

S ::= <expression> ";" <expression> ::= <item> <expression> ::= <item>{"+"|"-"}<expression> <item> ::= <factor> <item> ::= <factor>{"*"|"/"}<item> <factor> ::= <identifier> <factor> ::= <dec-const> <factor> ::= "("<expression>")"

The second group is to be used by a lexer:

<identifier> ::= <letter>[<letter>|<dec-digit>]... <dec-const> ::= [<dec-digit>]... <letter> ::= {'_'|'A'|'B'|...|'Z'|'a'|'b'|...|'z'} <dec-digit> ::= {'0'|'1'|'2'|'3'|'4'|'5'|'6'|'7'|'8'|'9'}

Syntax diagrams for the nonterminal symbol <expression> (as well as for other nonterminals) which are equivalent to the described production rules are shown in Figure 1.

As we can see, we defined a right-recursive grammar, hence it is possible to construct a recursive descent parser. To students’ attention one should mention here that, principally, a left-recursive grammar would fit the task better, since it produces right execution order for non-symmetrical operations like subtraction or division. However a left-recursive grammar cannot be



processed by a recursive descent parser (but may be transformed to an equivalent grammar with no left recursive rules [12]): the recursive non-terminal function call precedes the operation processing. So we use an alternative approach: right-recursive rules are used in combination with reading tokens in reverse order. In this case the only rule affected by such a reversion is the rule for parentheses embraced nested expression – a right parenthesis is discovered prior to a left one.

FIGURE 1: EXPRESSION GRAMMAR RULES.

III. CONSTRUCTING AN EXPRESSION PARSER: THE OLD TASK REVISITED In the domain of compiler and programming education much attention is paid to problems of

discovering simple but authentic models which might be perceivable by learners but doesn’t go against generality too much [13, 14, 15]. Thus, the old task of constructing an expression parser can still serve as a good example to be used in order to introduce lexical and syntactic analysis data models and type control structures coordinated with representing of the parsing process stages visually.

A. Lexical Parser Information Model and Bitwise Operations

The first problem we solve while constructing a lexical parser is how to recognize whether the input symbol does correspond to a set of symbols used in the common lexical context (e.g. sets of letters, digits, operators, punctuation signs, etc.). Hereafter we use the term “synterm” for such kind of sets. The solution isn’t as obvious as it’s usually believed. As far as 1976 Wirth described using sets and set operations in order to recognize synterms [16]. Manipulations with set abstractions fit most common cases (especially if synterms correspond to contiguous ranges like letters from 'A' to 'V' or digits from '0' to '9', etc.). However, if we consider more complex input languages (with paying attention to performance issues), set operations may be time-consuming. That’s why we propose a more general approach making possible to deal with various and irregular combinations of symbols used in the similar lexical context.

In fact, synterms correspond to the grammar symbols defined directly by using grammar terminals. For a set of synterms consisting of only one symbol we use the term of single-character synterm, otherwise we designate the respective set as a multi-character synterm. Some synterms



correspond to intersecting sets of symbols. Hereafter we mentioned them as intersecting synterms. In our example the intersecting multi-character synterms are the following:

“Letter or decimal digit” (LD10) corresponding to symbols that are allowed to be used while constructing identifiers;

“Decimal digit” (D10) corresponding to symbols used while constructing decimal literal values.

For a case of more complex input grammars there may be more intersecting synterms, for example, hexadecimal digits, letters, symbols used as numeric prefixes (like 'O', 'X'), etc. Operator symbols (like '+' or '–'), parentheses, and a semicolon are examples of single-character synterms which are non-intersecting for the input grammar E.

Formally, a finite set of synterms 푆 is a finite set of all subsets of 푉 . In the real world reasonably less multi-character synterms are required. For example, even to create a C++ code syntax analyzer, there is nothing more than about two dozen of them. Thus, for the required set of synterms 푆 :푆 ⊆ 푆 , |푆 | ≪ |푆 |.

Now lets us classify nonterminals of the grammar E in two groups:

A finite set of nonterminals 푉 defined by using synterms directly, i.e. 푉 ⊆ 푉 , such as 푉 ∷= 푃 (푉 ), 푉 ⊆ 푆 where 푃 (푉 ) means that the symbol 푉 is in the right part of the production rule 푃 , where 푃 ⊆ 푃.

Single characters which are finite sets 푆 from 푆 , such as 푆 = 1.

These two groups form a finite set of token types that a lexer recognizes. The solution corresponds to a lexer/parser cooperation model when a lexer is considered to be a separate component producing tokens which are further accepted by a parser, unlike to a scannerless model where tokenization and parsing are performed in a single step. Thus, for the grammar E this set of tokens includes nonterminals <identifier> and <dec-const>, and a set of single character tokens like ';', '+', '–', '*', '/', '(', ')'.

This formal model may be implemented by using specially defined synterm values. Particularly, in C++ the set of token types can be defined by using enumeration constants initialized explicitly (see Listing 1). Let us note that for the intersecting synterms binary representation codes with only one on-bit are used. The fragment in Listing 1 is suitable for the example limited by the expression grammar defined in section II. However let us note that the code is written in a manner to be easily adopted to “general” cases requiring additional synterm type constants.

The ONLY mark (which sets the highest synterm value bit to 1) is used to differentiate between single-character synterms and combined synterms. The latter are computed by using bitwise logical disjunctions of synterm values. The C++ language Synterm class serves as an example of both single-character and combined synterm definitions (see Listing 2).



LISTING 1: SYNTERM TYPES. enum SyntermTypes { // Intersecting synterms // In regards to identifiers: LAT = 0x0001, // LATin letter LD10 = 0x0002, // Latin or decimal Digit // In regards to numeric constants: D10 = 0x0010, // Digit_10 : decimal digit D0 = 0x0040, // Digit 0 D1 = 0x0080, // Digit 1 // In regards to non-intersecting // single character synterms: NOALP = 0x0000, // NOn-ALPhabetic : // out of language set ONLY = 0x8000, // ONLY : // non-intersecting synterm mark BLANK = (ONLY | ' '), // BLANK character // (space) LF = (ONLY | '\n'), // Line Finish LPARENT = (ONLY | '('), // Left PARENThesis RPARENT = (ONLY | ')'), // Right PARENThesis NIL = (ONLY | '\0') // Also used without special // named constant definition: // add ONLY | '+' // substract ONLY | '-' // multiply ONLY | '*' // divide ONLY | '/' // semicolon ONLY | ';' }; /* * Synterms masks (bit patterns) * used to recognize intersecting synterms */ enum SyntermMasks { MASK_NONE = 0xFFFF, MASK_LAT = (ONLY | LAT), MASK_LD10 = (ONLY | LD10), MASK_D10 = (ONLY | D10), MASK_D0 = (ONLY | D0), MASK_D1 = (ONLY | D1) };

LISTING 2: SYNTERM CLASS. /* * synterm.h (continuation) */ typedef unsigned int TSynterm; class Synterm { static TSynterm syntermTable[256]; public: static TSynterm getSynterm( char ch ) {



return syntermTable[ ch ]; } }; /* * synterm.cpp */ #include "synterm.h" // Synterm table (fragments) TSynterm Synterm::syntermTable[256] = { // Control symbols (no graphical form) /* ..0 0x00 ^@ NUL */ NIL, /* ..1 0x01 ^A SOH */ BLANK, /* ..2 0x02 ^B STX */ BLANK, /* ..3 0x03 ^C ETX */ BLANK, //... // Symbols with graphical form /* .32 0x20 ' ' */ BLANK, /* .33 0x21 '!' */ NOALP, /* .34 0x22 '\"' */ NOALP, /* .35 0x23 '#' */ NOALP, /* .36 0x24 '$' */ NOALP, /* .37 0x25 '%' */ NOALP, /* .38 0x26 '&' */ NOALP, /* .39 0x27 '\'' */ NOALP, // Single-charachter synterms /* .40 0x28 '(' */ LPARENT, /* .41 0x29 ')' */ RPARENT, /* .42 0x2A '*' */ ONLY | '*', /* .43 0x2B '+' */ ONLY | '+', // ... // Combined synterms /* .48 0x30 '0' */ D10 | LD10, /* .49 0x31 '1' */ D10 | LD10, /* .50 0x32 '2' */ D10 | LD10, // ... /* .57 0x39 '9' */ D10 | LD10, /* .58 0x3A ':' */ NOALP, /* .59 0x3B ';' */ ONLY | ';', /* .60 0x3C '<' */ NOALP, /* .61 0x3D '=' */ NOALP, /* .62 0x3E '>' */ NOALP, /* .63 0x3F '?' */ NOALP, /* .64 0x40 '@' */ NOALP, /* .65 0x41 'A' */ LAT | LD10, /* .66 0x42 'B' */ LAT | LD10, // ... /* .90 0x5A 'Z' */ LAT | LD10, /* .91 0x5B '[' */ NOALP, /* .92 0x5C '\\' */ NOALP, /* .93 0x5D ']' */ NOALP, /* .94 0x5E '^' */ NOALP, /* .95 0x5F '_' */ LAT | LD10, /* .96 0x60 '`' */ NOALP, /* .97 0x61 'a' */ LAT | LD10, /* .98 0x62 'b' */ LAT | LD10, // ...



/* 122 0x7A 'z' */ LAT | LD10, // ... /* 255 0xFF ' ' */ NOALP };

Combined synterm composition and recognition is explained in graphics in Figure 2 (by example of the Latin symbol 'A'). Note how the ONLY mark used in MASK_LAT allows differentiating between single-character synterms and the intersecting combined synterms. If a single-character synterm even has an on-bit which is the same that the on-bit of LAT synterm, the result of bitwise 'and' operation isn’t equal to LAT since the ONLY bit isn’t masked in this case (as you see by the example of the synterm RPARENT recognition).

Saying formally, the described model “inverts” the well known data structure where a finite set is represented by a binary vector [17]: binary synterm codes are used to be associated to symbol, so as to describe that a certain symbol belongs to a certain synterm. The information retrieval experts may find here an analogy with inverted indices used in document similarity and relatedness evaluation.

Litera type provides some syntactic sugar to deal with synterms while implementing lexer classes (see Listing 3). In Listing 3 the condition() method is a language level implementation of the process shown in Figure 2.

FIGURE 2: SYNTERM RECOGNITION.

LISTING 3: LITERA. /* * litera.h */ #include "synterm.h" struct Litera { char value; TSynterm synterm;



void set( char val ) { value = val; synterm = Synterm::getSynterm( val ); } bool condition( TSynterm check, TSynterm mask=MASK_NONE ) const { return (synterm & mask) == check; } bool isLAT() const { return condition( LAT, MASK_LAT ); } bool isD10() const { return condition( D10, MASK_D10 ); } bool isLD10() const { return condition( LD10, MASK_LD10 ); } bool isNil() const { return condition( NIL ); } bool isOnly() const { return condition( ONLY, ONLY ); } };

Now let’s summarize what have been explained before. Extracted tokens are to be constructed with use of information about intersecting sets of symbols used in the common lexical context. A synterm concept is introduced. Combined synterms are computed with using bitwise operations on binary values. Table of synterms is used while scanning input symbols. The method we use in order to code and to recognize synterms got from the input stream decreases number of conditions to be checked during lexical analysis and simplifies operations used in conditional statements. The model is extensible and configurable to fit more complex languages.

B. Lexer Structure and Finite State Machines

After all the synterms and synterm bitwise operations are defined, we are ready to discuss the lexer structure. Figure 3 illustrates the process of scanning an arithmetic expression with constructing identifiers, decimal literals and single-character tokens. We use a graphical notation inspired by the software development visual formalism introduced in [18] to represent lexer structure. The essentials of the class hierarchy representing token types recognized by the lexer are represented in Listing 4.

Token builders implement methods the lexer calls while constructing multi-character tokens. Token recognition problem is a suitable area to introduce a concept of a finite state machine (FSM) as Figure 4 shows.



Since the multi-character token construction has structural similarity for different tokens, there is a space to define a token builder abstract model. Token builder interface includes three basic operations at least:

FIGURE 3: LEXER STRUCTURE.

FIGURE 4: IDENTIFIER FSM STATECHART.

lexFirst() method is being called when we have got the first token symbol; lexNext() method is being called at every iteration while subsequent token characters



are being scanned; lexFinish() method completes the token construction and returns the token.

The corresponding abstract class may be defined as Listing 5 illustrates.

LISTING 4: C++ TOKEN CLASS HIERARCHY. /* * token.h */ #include <sstream> #include <string> using namespace std; enum TokenType { LT_NONE, LT_IDENT, LT_DECCONST, LT_OPERATION, LT_SINGLE }; enum OperationType { OP_PLUS, OP_MINUS, OP_MULT, OP_DIVIDE, OP_END }; class Token { public: TokenType type; virtual string toString() const = 0; }; template <class TokenData> class BasicToken : public Token { public: TokenData value; // General implementation may fits // many token types // (may be overridden) virtual string toString() const { string buffer; ostringstream os( buffer ); os << value; return os.str(); } }; class Numeric : public BasicToken<int> {}; class Ident : public BasicToken<string> {}; class Operation : public BasicToken<OperationType> { public: string toString() const { switch( value )



{ case OP_PLUS: return string("+"); break; //... case OP_MINUS: return string("-"); break; //... } } }; class SingleCharacter : public BasicToken<char> {}; LISTING 5: C++ ABSTRACT TOKEN BUILDER. /* * lexbuild.h */ class AbstractLexBuilder { public: virtual void lexFirst( char ) = 0; virtual void lexNext( char ) = 0; virtual Token *lexFinish() = 0; };

The identifier builder and the decimal literal builder have to implement the AbstractLexBuilder interface as we show in Appendix A (Listing 8) by the example of the identifier construction process which corresponds to the model in Figure 4.

Thus, introducing the lexer structure gives space for a discussion about inheritance. A hierarchy of token classes sharing the common interface of Token class give space to the explanation of internal data representation particularities in derived classes Ident, Numeric, Operation, and SingleCharacter.

Especially in C++, there is an illustration of combined usage of such concepts as template based parametric polymorphism (class BasicToken) and dynamic binding.

The source code of Lexer class is presented in Appendix B (Listing 9). Since we are favor for literate programming ideas introduced by Knuth as far as 1984 [19], we believe the source code to be readable as is and skip additional explanations.

C. Implementing Grammar Rules: A Recursive Hierarchy

Despite the fact that a classical FSM model implicitly assumes input forwarding at every transition, in the software implementation (which is not limited by FSM restrictions) it is more practical to have explicit separate lexer operations for getting an input symbol and forwarding to the next symbol from the input sequence. Let us illustrate the possible implementation of the hierarchy of functions used while parsing an expression. The functional structure corresponds strongly to syntax diagrams represented in Figure 1, namely: each grammar nonterminal



matches the (non-terminal) function.

The C++ abstract parser definition together with the Polish notation container class is shown in Listing 6.

LISTING 6: ABSTRACT PARSER CLASS. /* * parser.h */ // Polish notation class class Polish { protected: vector<Token*> *body; public: Polish (); void put( Token* token ); vector<Token*> *getVector (); }; // Parser class class Parser { protected: Polish polish; public: virtual void parse() = 0; virtual vector<Token*> *getResult() { return polish.getVector(); } };

Now it’s time to implement a parser for the grammar defined in section 2. The C++

ReverseParser class (presented in Listing 7) contains methods corresponding to non-terminal functions for the expression grammar E production rules. To illustrate the idea it is enough to expose the parse() and expression() method definitions.

LISTING 7: PARSER CLASS (FRAGMENT). /* * reverseparser.h */ #include "parser.h" class ReverseParser : public Parser { Lexer lexer; void expression(); void item(); void factor(); public: ReverseParser( char *inputString, IdTable *idTable ) : lexer( inputString, idTable ){}



void parse(); }; /* * reverseparser.cpp */ #include "reverseparser.h" void ReverseParser::parse() { lexer.initReverse(); expression(); Token *token = lexer.get(); Operation *current = dynamic_cast<Operation*>(token); if( token->type==LT_OPERATION && current->value==OP_END ) { polish.put(token); return; } throw ParserException(); } void ReverseParser::expression() { item(); Token *token = lexer.get(); switch( token->type ) { case LT_OPERATION: Operation *current = dynamic_cast<Operation*>(token); switch( current->value ) { case OP_PLUS: case OP_MINUS: lexer.next(); expression(); polish.put(token); break; } } }

Extensible and pure object-oriented model with use of dynamic casts may have performance drawbacks if there are many runtime type conversions. An instructor may introduce here a more efficient choice based on multiple dispatch and type switching [20, 21, 22].

To sum up, we obtain a recursive parser structure similar to the model introduced by Gries in [3], but implemented in object-oriented style. Figure 6 represents the recursive parser structure in a whole. Dashed lines correspond to the recursive method calls. Students are welcome to discover evident similarity between the program structure and grammar diagrams, so we can name the code constructed in such a manner as a source code based grammar definition. Note that the actions which relate to the operations on reverse Polish notation are in fact parts of code generation component (we show them grayed in Figure 5).



D. Expression Computation and a Concept of Computational Stack

After we get a polish notation representation, we are able to introduce the computation procedure and a concept of computation stack. Let us note that at this point there is also a bridge to a syntax tree representation of the scanned expression.

E. Assembling a Jigsaw Puzzle

Figure 6 represents the system architecture of our solution. As it is mentioned in [23], one of difficulties in learning inheritance and polymorphism concepts is that students doesn’t understand well class hierarchies and their role in inheritance, so the practically oriented example may help learners in accepting inheritance ideas and commonness, but also in analysis of inheritance based type control drawbacks and limitations.

FIGURE 5: EXPRESSION RECURSIVE PARSER.



FIGURE 6: EXPRESSION PARSER SYSTEM ARCHITECTURE.

IV. CONCLUSION The paper introduces the problem of parenthesis-free arithmetic expression parsing and

computation as a type of end-to-end task which encompasses some important models and formalisms of computer science and programming. Table I summarizes this information.

The solution requires a combination of computer science concepts and programming techniques and mechanisms presentable within the framework of one problem. The problem is simple enough to meet the needs of an academic course, at the same time it has logical and structural complexity.

We believe that due to visual representations used at every step the solution complexity decreases; thereby students are expected to be capable to create a clear mental model of the software structure as well as of its execution. Following Milne and Rowe (see [13]), absence of such a mental model is fairly considered as one of usual difficulties in learning programming.

Another aspect that we tried to illustrate in this paper is bridging the gap between content



concepts (e.g. data structures) and process concepts (e.g. “grammar driven” recursive computation) [24].

As a matter of things, this article may seems to be a collection of lecture notes for several lecture hours of a practical course on programming. It is so. Furthermore, a sequence of multi-aspect tasks presented in a way used in this material may be considered as a framework for a special problem-oriented course on software technology.

TABLE I: PARENTHESIS-FREE EXPRESSION PARSING AS A MULTI-ASPECT TASK

Solution stage Underlying models

Programming language focus

Visual models In-depth focus

Requirement definition

Recursive grammar

Unit tests planning

Syntactic diagrams

Left- and right- recursive grammars, grammar transformation

Synterm definition

Sets, bitwise operations

Enumerations, symbol tables

Masked bitwise operations

Adopting a synterm recognition model to more complex languages

Lexical analysis

Finite state machine, hash search

Inheritance, dynamic binding, parametric polymorphism

Visual formalisms, state charts, class diagrams

Upcasting, downcasting, run time type information, regular expressions

Syntactic analysis

Recursive descent parser

Recursive functions

Syntax trees, class digrams

Type switch, multiple dispatch

Code generation

Polish notation Collections Collection interfaces

Computation (execution)

Stack Stack operations

Stack implementations, transforming recursion to iteration



APPENDIX A. IdentBuilder implementation source code (C++)

LISTING 8: IDENTIFIER BUILDER C++ IMPLEMENTATION. /* * lexbuild.h (continuation) */ class IdentBuilder : public AbstractLexBuilder { static const int Ident_Size; int ixLit; int limit; char value[ IDENT_SIZE ]; public: IdentBuilder() { value[0]='\0'; } void lexFirst( char ); void lexNext( char ); Token *lexFinish(); }; /* * lexbuild.cpp */ #include "lexbuild.h" const int IdentBuilder::Ident_Size = IDENT_SIZE; void IdentBuilder::lexFirst( char ch ) { ixLit = 0; limit = Ident_Size; value[0] = ch; } void IdentBuilder::lexNext( char ch ) { if( ixLit >= limit ) return; ++ixLit; value[ixLit] = ch; } Token *IdentBuilder::lexFinish() { ++ixLit; value[ixLit] = '\0'; Ident *ident = new Ident(); ident->value = string( value ); ident->type = LT_IDENT; return ident; }



B. Lexer implementation source code (C++)

LISTING 9: LEXER CLASS AND ITS IMPLEMENTATION. /* * lexer.h */ #include <vector> using namespace std; #include "synterm.h" #include "litera.h" #include "lexbuild.h" #include "idtable.h" #include "token.h" class LexerException{}; class Lexer { char *inputString; char *ptrInputString; Token* currentToken; vector<Token*> *input; vector<Token*>::iterator itInput; IdTable *idTable; void readInputString(); void addToken( Token* ); TSynterm getlit(Litera&); void ungetlit(); // may be needed for // some implementations public: Lexer( char* inputString, IdTable *idTable ) { ptrInputString = this->inputString = inputString; this->idTable = idTable; input = new vector<Token*>; readInputString(); itInput = input->begin(); } void initStraight(); void initReverse(); vector<Token*> *getInputTokens(); Token* get(); Token* getAndNext(); void unget(); void next(); private: void readInput(); };



/* * lexer.cpp */ #include "lexer.h" TSynterm Lexer::getlit(Litera& litera) { litera.set( *ptrInputString ); if( litera.synterm == NIL ) return NIL; ++ptrInputString; return litera.synterm; } void Lexer::ungetlit() { if( ptrInputString == inputString ) throw LexerException(); --ptrInputString; } void Lexer::readInputString() { Litera litera; for(;;) { if( getlit(litera) == NIL ) throw LexerException(); // Missing spaces while( litera.synterm == BLANK ) { getlit( litera ); } if( litera.isNil()) throw LexerException(); if( litera.isLAT() ) { IdentBuilder *identBuilder = new IdentBuilder(); identBuilder->lexFirst( litera.value ); for(;;) { if( getlit(litera)==NIL ) { break; } if( litera.isLD10() ) { identBuilder->lexNext( litera.value ); } else { ungetlit(); break; } } currentToken = identBuilder->lexFinish(); addToken( currentToken );



delete identBuilder; continue; } else if( litera.isD10() ) { DecConstBuilder *decBuilder = new DecConstBuilder(); decBuilder->lexFirst( litera.value ); for(;;) { if( getlit(litera)==NIL ) { break; } if( litera.isD10() ) { decBuilder->lexNext( litera.value ); } else { ungetlit(); break; } } currentToken = decBuilder->lexFinish(); addToken( currentToken ); continue; } else if (litera.isOnly() ) { Operation *operToken = new Operation(); switch(litera.synterm) { case ONLY | '+': operToken->type = LT_OPERATION; operToken->value = OP_PLUS; addToken(operToken); break; case ONLY | '-': operToken->type = LT_OPERATION; operToken->value = OP_MINUS; addToken(operToken); break; case ONLY | '*': operToken->type = LT_OPERATION; operToken->value = OP_MULT; addToken(operToken); break; case ONLY | '/': operToken->type = LT_OPERATION; operToken->value = OP_DIVIDE; addToken(operToken); break; case ONLY | ';': operToken->type = LT_OPERATION; operToken->value = OP_END; addToken(operToken); return; } if( operToken->type != LT_OPERATION ) { SingleCharacter *singleCharToken = new SingleCharacter(); switch( litera.synterm ) {



case LPARENT: singleCharToken->type = LT_SINGLE; singleCharToken->value = '('; addToken(singleCharToken); break; case RPARENT: singleCharToken->type = LT_SINGLE; singleCharToken->value = ')'; addToken(singleCharToken); break; default: throw LexerException(); } } } else throw LexerException(); } } void Lexer::addToken( Token* token ) { input->push_back( token ); if( token->type == LT_IDENT ) { idTable->add( token->toString().c_str()); } } void Lexer::initStraight() { itInput = input->begin(); } void Lexer::initReverse() { int size = input->size()-1; for (int i=0; i<size/2; i++ ) { Token* copy = (*input)[i]; (*input)[i] = (*input)[size-1-i]; (*input)[size-1-i]=copy; } itInput = input->begin(); } Token* Lexer::get() { return *itInput; } Token* Lexer::getAndNext() { return *itInput++; } void Lexer::unget() { itInput--; } void Lexer::next() { ++itInput; }

ACKNOWLEDGMENT I thank my colleague Dr. Mikhail Glukhikh from St. Petersburg Polytechnic University for his

generous help in teaching the undergraduate course of programming and his kind assistance in preparing examples.



REFERENCES [1] Shneiderman, B. (1980). Software Psychology: Human Factors in Computer and Information Systems.

Winthrop Publishers.

[2] Burks, A. W., Warren, D. W., Wright, J. B. (1954). An analysis of a logical machine using parenthesis-free notation. Mathematical Tables and Other Aids to Computation, 8(46), 53-57.

[3] Gries, D. (1971). Compiler Construction for Digital Computers. John Wiley & Sons, Inc., New York, NJ, USA.

[4] Dahl, E. W., Dijkstra, E., Hoare, C. A. R., editors. (1972). Structured Programming. Academic Press Ltd., London, UK.

[5] Foster, J. M. (1970). Automatic syntactic analysis. Computer monographs. Mac Donald New York, London.

[6] Weingarten, F. W. (1973). Translation of computer languages. Holden-Day computer and information sciences series. Holden-Day, San Francisco, London, Toronto.

[7] Ward, S. A. & Halstead Jr., R. H. (1990) Computation Structures. MIT electrical engineering and computer science series. MIT Press.

[8] Iwama, F., Nakamura, T., Takeuchi, H. (2012). Constructing parser for industrial software specifications containing formal and natural language description. In Proceedings of the 34th International Conference on Software Engineering (ICSE), June 2012, 1012-1021.

[9] Aho, A. V. & Ullman, J. D. (1972). The Theory of Parsing, Translation, and Compiling. Prentice-Hall, Inc., Upper Saddle River, NJ, USA.

[10] Sommerville, I. (2010). Software Engineering (9th ed.). Addison-Wesley, Harlow, England.

[11] Pyshkin, E. (2011). Teaching programming: What we miss in academia. In 7th Central and Eastern European Software Engineering Conference in Russia (CEE-SECR), pages 1-6. IEEE.

[12] Greibach, S. A. (1964). Formal parsing systems. Commun. ACM, 7(8):499-504.

[13] Milne, I. & Rowe, G. (2002). Difficulties in learning and teaching programming views of students and tutors. Education and Information Technologies, 7(1), 55-66.

[14] Urquiza-Fuentes, J., Manso, F., Velázquez-Iturbide, J. Á., Rubio-Sánchez, M. (2011). Improving compilers education through symbol tables animations. In Proceedings of the 16th Annual Joint Conference on Innovation and Technology in Computer Science Education, ITiCSE ‘11, 203-207, New York, NY, USA.

[15] Mahoney, M. & Pedersen, J. (2010). Teaching compiler code generation: Simpler is better. SIGCSE Bull., 41(4), 30-34.

[16] Wirth, N. (1978). Algorithms + Data Structures = Programs. Prentice Hall PTR, Upper Saddle River, NJ, USA.

[17] Aho, A. V., Hopcroft, J. E., Ullman, J. (1983). Data Structures and Algorithms (1st ed.). Addison-Wesley Longman Publishing Co., Inc., Boston, MA, USA.



[18] Lekarev, M. F. (1993). Das graphische Verfahren der Software-Entwicklung für logisch komplizierte Anwendungen. In Fachhoshschule Hamburg Tech. Bericht.

[19] Knuth, D. E. (1984). Literate programming. The Computer Journal, 27(2), 97-111.

[20] Shopyrin, D. (2006). Multimethods implementation in C++ using recursive deferred dispatching. IEEE Software, 23(3), 62-73.

[21] Muschevici, R., Potanin, A., Tempero, E., Noble, J. (2008). Multiple dispatch in practice. In Proceedings of the 23rd ACM SIGPLAN Conference on Object-oriented Programming Systems Languages and Applications, OOPSLA ‘08, 563-582, New York, NY, USA.

[22] Solodkyy, Yu., Dos Reis, G., Stroustrup, B. (2012). Open and efficient type switch for C++. SIGPLAN Not., 47(10), 963-982.

[23] Liberman, N., Beeri, C., Kolikant, Y. B-D. (2011). Difficulties in learning inheritance and polymorphism. Trans. Comput. Educ., 11(1):4:1-4:23.

[24] Zendler, A., McClung, O. W., Klaudt, D. (2012). Content and process concepts relevant to computer science education: A cross-cultural study. International Journal of Research Studies in Computing, 1(2).

AUTHOR’S BIOGRAPHY

Evgeny Pyshkin received a Ph.D. degree in Computer Science from St. Petersburg State Polytechnic University (Russia) in 2000. He is presently a senior associate professor of the Institute of Computing and Control at the same university. He is interested in developing educational programs and research in areas of software engineering, information retrieval and data processing. His special interests are music information retrieval systems, object-oriented design and software testing automation. Dr. Pyshkin is an author of several books on programming. He is a program committee member of several international

conferences sponsored by ACM, FTRA and IEEE. As an invited professor he taught undergraduate and graduate school courses on programming, software development frameworks and software engineering for students of the Central Ostrobothnia University (Finland) and the University of Aizu (Japan).


Structured Stream Data Mining Using SVM Method as Basic Classifier

Authors

Hadi Barani Baravati Department of Computer Engineering/ Islamic Azad University, Iranshahr Branch, Iran

[email protected] Iranshahr, Iran

Javad Hosseinkhani Department of Computer Engineering/ Islamic Azad University, Science and Research Branch, Iran


Solmaz Keikhaee Department of Electrical Engineering/ Islamic Azad University, Science and Research Branch, Iran

Abbas Shahsavari Department of Computer, Qeshm International Branch, Islamic Azad University, Iran


[email protected] Qeshm Iran

Abstract

Recently, the huge number of email spams has caused serious problems in essential email communication. In this paper, we describe the results of an empirical study on one spam detection method namely Support Vector Machines (SVMs). To conduct the study, first the recieved emails would be pre proccessed then stream data in order to learning the classification would be given to the proposed data miner system. The number of training data set with window based solution will be selected with default , W=100 , the first 100 data would be used as trainig set. each receieved email input to SVM to be classified in to 2 predefined classes named: Non spam, and Spam. A program is written that 4 different kinds of time window in order to SVM training are selected (100,200,500 and all the preset data or open window). The evaluation criteria include accuracy rate, recall, and precision rate. The results indicate that the approach has its pros and cons.

Key Words

Spam Detection, Email Classification, Support Vector Machine.

Structured Stream Data Mining Using SVM Method as Basic Classifier Hadi Barani Baravati, Javad Hosseinkhani, Solmaz keikhaee, and Javid Hosseinkhani Naniz


I. INTRODUCTION Most Internet users use mail to communicate electronically. They depend on the mail system to deliver their mails to the recipient. Spam has made the mail system more unreliable because mail can get falsely caught by spam filters on the way to the recipient or mail can drown among spam in the recipient’s inbox. The goal of the Internet community should be to work toward a more usable Internet with less spam. Possible ways to do this are through the law and the legal system, technical solutions and user awareness. Recently, the huge number of email spams has caused serious problems in essential email communication. Traditional spam filters aim at analyzing email content to characterize the features that are commonly included in spams. However, it is observed that crafty tricks designed to avoid content-based filters will be endless owing to the economic benefits of sending spams. In view of this situation, there has been much research effort toward doing spam detection based on the reputation of senders rather than what is contained in emails. Motivated by the fact that spammers are prone to have unusual behavior and specific patterns of email communication, exploring email social networks to detect spams has received much attention [19]. Email communication is prevalent and indispensable nowadays. However, the threat of unsolicited junk emails, also known as spams, is increasingly serious. According to a survey by the website Top Ten REVIEWS [19], 40% (12.4 billion out of 31 billion per day) of emails were considered as spams in 2006. The statistics collected by Message- Labs show that spam rate persistently remains high. The primary challenge for spam detection problem lies in the fact that for the purposes of gaining economic benefits, distributing spyware, and spreading links to phishing websites, to name a few, spammers will always develop new sophisticated approaches to attack spam filters. For example, traditional text-based filters, such as Naive Bayes classifiers, have been commonly passed by obfuscated keywords and random paragraph insertion. In addition to essential information that spammers want to convey, there are various unrelated contents included in spams. Since unexpected tricks employed in email content are ceaseless, a number of studies have focused on identifying who sends the email rather than what is contained in the email. On-line data stream mining has attracted much research interest, but systems that can be used as a workbench for online mining have not been researched, since they pose many difficult research challenges [18]. On-line data stream mining plays a key role in growing number of real-world applications, including network traffic monitoring, intrusion detection, web click-stream analysis, and credit card fraud detection. Thus, many research projects have recently focused on designing fast mining algorithms, whereby massive data streams can be mined with real-time response [10, 8, 9, and 7]. Similarly, many research projects have also focused on managing the data streams generated from these applications [11, 12, and 13]. However, the problem of



www.elvedit.com 308

supporting mining algorithms in such systems has, so far, not received much research attention [14]. This situation seems unusual, since the need for a mining system for static data mining, was immediately recognized [17] and has lead to systems such as, Weka [15] and OLE DB for DM [16]. Furthermore, static mining algorithms can also be written in procedural language using a cache mining approach that makes little use of DBMS essentials. However, online mining tasks cannot be deployed as stand-alone algorithms, since they require many DSMS essentials, such as I/O buffering, windows, synopses, load shedding, etc. Clearly, KDD researchers and practitioners would rather concentrate on the complexities of data mining tasks and avoid the complexities of managing data streams, by letting the mining system handle them. In short, while mining systems are a matter of convenience for stored data, they are a matter of critical necessity for data streams.

A method of ordering linear and nonlinear data is Support vector machines (SVMs). In a case, SVM is an algorithm and the function of it is as follows. To change the original training data into a higher dimension, it applies a nonlinear mapping. It seeks for the linear ideal separating hyper plane through this new dimension. A hyper plane can always separate the data into two classes with a suitable nonlinear mapping to an appropriately high dimension. The SVM discovers this hyper plane utilizing support vectors that is “essential” training tuples and margins which is explained by the support vectors. Vladimir Vapnik and colleagues Isabelle Guyon and Bernhard Boser (1992) have done the first research on support vector machines since the groundwork for SVMs has been around. Even though the training time of SVMs is very extremely slow, they are very precise and can to model compound nonlinear decision limitations. In compare to other methods, they are much less predisposed to over fitting. The provision vectors also are a compressed explanation of the trained model. SVMs also are able to utilize for numeric calculation along with classification. They have been used for many areas such spam email detection. [6].

There are many anti-spam strategies and methods. In this paper, we describe the results of an empirical study on one spam detection method namely Support Vector Machines (SVMs). The reason for choosing this method is that it has good theoretical foundation, scale up well with large data, and lend itself to the text classification problem. In our study, we implemented an application of SVMs. To conduct the study, first the recieved emails would be pre proccessed then stream data in order to learning the classification would be given to the proposed data miner system. The number of training data set with window based solution will be selected with default , W=100 , the first 100 data would be used as trainig set. each receieved email input to SVM to be classified in to 2 predefined classes named: Non spam, and Spam. A program is written that 4 different kinds of time window in order to SVM training are selected (100,200,500 and all the preset data or open window). The evaluation criteria include accuracy rate, recall, and precision rate. The results indicate that the approach has it pros and cons.



II. RELATED WORKS Since the email spam problem is more and more serious nowadays, various techniques have been explored to relieve this problem. According to what features of emails are being used, previous works on spam detection can be generally classified into three categories: (1) content-based methods, (2) non-content-based methods, and (3) integrated methods. Initially, researchers analyze email content text and model this problem as a binary text classification task. Representatives of this category are Naive Bayes [20, 21] and Support Vector Machines (SVMs) [22, 23] methods. In general, Naive Bayes methods train a probability model using classified emails, and each word in emails will be given a probability of being a suspicious spam keyword. As for SVMs, it is a supervised learning method, which possesses outstanding performance on text classification tasks. Traditional SVMs [23] and improved SVMs [22] have been investigated. While above conventional machine learning techniques have reported excellent results with static data sets, one major disadvantage is that it is cost prohibitive to constantly re-train these methods with the latest information to adapt to the rapid evolving nature of spams. Moreover, crafty content obfuscation tricks have always been developed to degrade the performance of these approaches. On the other hand, certain specific features such as URLs [24] and images [25] are also taken into account for spam detection. The other group attempts to exploit non-content information such as email header, email traffic [26], and email social network [27, 28] to filter spams. Collecting notorious and innocent sender addresses (or IP addresses) from email header to create black list and white list is a commonly applied method initially. In [26], the authors intend to analyze email traffic flows to detect suspicious machines and abnormal email communication. It is noted that these approaches have to operate in coordination with other complementary methods to gain better results. Moreover, in [29], a pure reputation system is designed to apply in a large webmail service. This system is constructed by the past behavior of each sender with SPF and DomainKey authentication.

Furthermore, some researchers consider combining the merits of several techniques [30, 31, 32]. Even though the performance of classifier integration seems prominent, there is still no conclusion on what is the best combination. In addition, how to efficiently update the whole included classifiers is another unsolved issue. In [33], certain network related features are extracted to characterize each user. A modified k-Nearest Neighbor (k-NN) model is then employed to perform the spam classification. In [34], graph theoretical analysis of networks is presented to discover good discriminators between legitimate emails and spams. In [35], the authors propose an email scoring mechanism that infers reputation ratings between individuals in networks. In [27], the authors exploit the feature of clustering coefficient in networks to devise a detection mechanism. Overall, these works generally suffer from the following two problems. First, they are not robust in diverse



www.elvedit.com 310

environments. The other is that the update scheme, which is critical for evolving networks, has been ignored in these works.

III. SUPPORT VECTOR MACHINES We consider spam detection as a text classification problem. There are two classes for email messages: 풚풊€ {-1, +l} where -1 indicates no spam and +1 spam. A feature is a word in an email message and a feature vector 풙풊 represents an email in the feature space. Given n labeled training examples: (풙ퟏ, 풚ퟏ), ..., (풙풏, 풚풏), the task is to learn from the training examples a hypothesis that can be used to classify unseen email messages. Support vector machines are a family of learning methods [l]. Linear hard-margin SVMs are the simplest model in SVMs and are also called the maximal margin classifier which works only for data linearly separable in the feature space. The linear hard-margin SVMs separate feature vectors into the two classes by finding a hyper plane with maximal margin. The feature vectors closest to the hyper plane are called support vectors. The maximal margin hyper plane bounds the generalization error of the linear machines given a training set S, and can be obtained by maximizing the function

W(α) = a −12

a a y y x . x

Subject to:

y a = 0, a ≥ 0.

Soft-margin SVMs [1] can be used for non-linearly separable data. Soft-margin SVMs allow training errors. The optimization problem now becomes maximizing the following:

W(α) = a −12

a a y y x . x



Subject to:

y a = 0, 0 ≤ a ≤ C.

The C is the parameter that we need to tune to make the model fit to the non-linearly separable data. The soft margin SVMs behave like hard-margin SVMs if the parameter C is large enough. See [2, 3, 4, and 5] for details.

IV. RESEARCH DESIGN AND PROPOSED FRAMEWORK According to Figure 1, first the recieved emails would be pre proccessed then stream data in order to learning the classification would be given to the proposed data miner system. The number of training data set with window based solution will be selected with default , W=100 , the first 100 data would be used as trainig set.

FIGURE 1: PROPOSED FRAMEWORK

Before the emails or Data are used for retrieval, some preprocessing tasks are usually performed. The tasks are stopword removal and stemming. Stopwords are frequently occurring and insignificant words in a language that help construct sentences but do not represent any content of the documents. Articles, prepositions and conjunctions and some pronouns are natural candidates. In many languages, a word has various syntactical forms depending on the contexts that it is used.

For example, in English, nouns have plural forms, verbs have gerund forms (by adding “ing”), and verbs used in the past tense are different from the present tense. These are considered as syntactic variations of the same root form. Such variations cause low recall for a detection system because a relevant spam email may contain a variation of a query word but not the exact word itself. This problem can be partially dealt with by stemming.



www.elvedit.com 312

FIGURE 2: CLASSIFICATION OF RECEIVED EMAILS TO SVM

Figure 2 illustrates that each receieved email input to SVM to be classified in to 2 predefined classes named: Non spam, and Spam. Data set Usnet1 and Usenet 2 are applied inroder to training and proposed data miner learning .

V. EXPERIMENTAL RESULT Table 1 shows the stream data classificiation exprimental results through using SVM in Spam data set. In this table, 4 different kinds of time window in order to SVM training are selected (100,200,500 and all the preset data or open window) that 3 evaluations criteria’s including precision , recall and accuracy are evaluated that is shown in table 1.

TABLE1: STREAM DATA CLASSIFICIATION EXPRIMENTAL RESULTS

Precision Recall Accuracy

Simple Incremental 0.9987 0.9172 0.9320

Time Windows(W=100) 0.9600 0.9165 0.9140

Time Windows(W=200) 0.9660 0.8954 0.9050

Time Windows(W=500) 0.9937 0.8275 0.8990

Figure 3 shows the mean of vector precision for 1000 experimental samples.



FIGURE III: THE MEAN OF VECTOR PRECISION FOR 1000 EXPERIMENTAL SAMPLES.

VI. CONCLUSION On-line data stream mining has attracted much research interest, but systems that can be used as a workbench for online mining have not been researched, since they pose many difficult research challenges. On-line data stream mining plays a key role in growing number of real-world applications, including network traffic monitoring, intrusion detection, web click-stream analysis, and credit card fraud detection. Thus, many research projects have recently focused on designing fast mining algorithms, whereby massive data streams can be mined with real-time response. In our experiments, the linear SVMs have several advantages. One important feature is that SVMs are less influenced by the sizes of training cases in the two classes because they are not geared toward minimizing the error rate, but instead attempt to separate the patterns in feature space. However, the performance is the potential issue. If there are large numbers of training cases, learning process can be long. Execution can be slow for nonlinear SVMs.

0 100 200 300 400 500 600 700 800 900 10000.88

0.9

0.92

0.94

0.96

0.98

1

Time

Acc

urac

y

0 100 200 300 400 500 600 700 800 900 10000.88

0.9

0.92

0.94

0.96

0.98

1

Time

Acc

urac

y

0 100 200 300 400 500 600 700 800 900 10000.82

0.84

0.86

0.88

0.9

0.92

0.94

0.96

0.98

1

Time

Acc

urac

y

0 100 200 300 400 500 600 700 800 900 10000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Time

Acc

urac

y



www.elvedit.com 314

To get more accurate results, such study needs to be repeated on larger data sets. It is also clear in our study that the performance of those classification methods really depend on the training examples, i.e. the feature vectors extracted from the original email messages. In the experiments, only words in the message body were used as the candidates of the features. More important information might be missed in the feature extraction process. For example, the subject title of email can be the good candidate of the features. And also, recent spam messages are coded with html so it might be a good idea to include the html codes in the features. The preprocessing before running the learners is the important phase for the learners to perform better classification. For the further analyses, the spam emails for training and testing should be decomposed into the multiple classes according to the kinds of spam emails such as investment and vacations. This makes the analysis results more useful and refined.

REFERENCES [1] Matsumoto, Ryota, Du Zhang, and Meiliu Lu. "Some empirical results on two spam detection

methods." Information Reuse and Integration, 2004. IRI 2004. Proceedings of the 2004 IEEE International Conference on. IEEE, 2004.

[2] T. Joachims, Learning To Classify Text Using Support Vector Machines, Kluwer Academic Publishers, 2002.

[3] E. Osuna, R. Freund, and E Girosi, Improved Training Algorithm for Support Vector Machines, Roc. IEEE NNSP'97, pp.276-285, 1997.

[4] J. C. Platt, Sequential Minimal Optimization: A Fast Algorithm for Training Support Vector Machines, Advances in Kernel Method: Support Vector Learning, by Scholkopf, Burges, and Smola, MIT Press, pp. 185- 208,1998.

[5] V. Vapnik, Estimation of Dependencies Based on Empirical Data, Springer-Verlag, New York, 1982.

[6] Moradi Koupaie, Hossein, Suhaimi Ibrahim, and Javad Hosseinkhani. "Outlier Detection in Stream Data by Machine Learning and Feature Selection Methods." International Journal of Advanced Computer Science and Information Technology (IJACSIT) Vol 2 (2014): 17-24.

[7] B. Mozafari, H. Thakkar, and C. Zaniolo. Verifying and mining frequent patterns from large windows over data streams. In ICDE, 2008.

[8] F. Chu and C. Zaniolo. Fast and light boosting for adaptive mining of data streams. In PAKDD, volume 3056, 2004.

[9] H. Wang and et al. Mining concept-drifting data streams using ensemble classifiers. In SIGKDD, 2003.

[10] George Forman. Tackling concept drift by temporal inductive transfer. In SIGIR, pages 252–259, 2006.

[11] Yan-Nei Law, Haixun Wang, and Carlo Zaniolo. Data models and query language for data streams. In VLDB, 2004.



[12] A. Arasu, S. Babu, and J.Widom. CQL: A language for continuous queries over streams and relations. In DBPL, 2003.

[13] D. Abadi et al. Aurora: A new model and architecture for data stream management. VLDB Journal, 2003.

[14] H. Thakkar, B. Mozafari, and C. Zaniolo. Designing an inductive data stream management system: the stream mill experiences. In Scalable Stream Processing Systems, 2008.

[15] Weka 3: data mining with open source machine learning software in java. http://www.cs.waikato.ac.nz.

[16] Z. Tang and et al. Building data mining solutions with OLE DB for DM and XML analysis. SIGMOD Record, 2005.

[17] Tomasz Imielinski and Heikki Mannila. A database perspective on knowledge discovery. Commun. ACM, 1996.

[18] Thakkar, Hetal, Barzan Mozafari, and Carlo Zaniolo. "A data stream mining system." Data Mining Workshops, 2008. ICDMW'08. IEEE International Conference on. IEEE, 2008.

[19] Tseng, Chi-Yao, and Ming-Syan Chen. "Incremental SVM model for spam detection on dynamic email social networks." Computational Science and Engineering, 2009. CSE'09. International Conference on. Vol. 4. IEEE, 2009.

[20] J. Hovold. Naive bayes spam filtering using wordposition- based attributes. Proc. of the 2nd Conference on Email and Anti-Spam (CEAS), 2005.

[21] V. Metsis, I. Androutsopoulos, and G. Paliouras. Spam filtering with naive bayes – which naive bayes Proc. of the 3rd Conference on Email and Anti-Spam (CEAS), 2006.

[22] E. Blanzieri and A. Bryl. Evaluation of the highest probability svm nearest neighbor classifier with variable relative error cost. Proc. of the 4th Conference on Email and Anti-Spam (CEAS), 2007.

[23] H. Drucker, D. Wu, and V. N. Vapnik. Support vector machines for spam categorization. Proc. of the IEEE Transactions on Neural Networks, pages 1048–1054, 1999.

[24] K. M. Schneider. Brightmail url filtering. Proc. of the MIT Spam Conference, 2004.

[25] M. Dredze, R. Gevaryahu, and A. Elias-Bachrach. Learning fast classifiers for image spam. Proc. Of the 4th Conference on Email and Anti-Spam (CEAS), 2007.

[26] R. Clayton. Email traffic: A quantitative snapshot. Proc. of the 4th Conference on Email and Anti-Spam (CEAS), 2007.

[27] P. O. Boykin and V. Roychowdhury. Leveraging social networks to fight spam. Proc. of the IEEE Computer, 2005.

[28] P.-A. Chirita, J. Diederich, and W. Nejdl. Mailrank: Using ranking for spam detection. Proc. of the 14th ACM International Conference on Information and Knowledge Management (CIKM), pages 373– 380, 2005.



www.elvedit.com 316

[29] B. Taylor. Sender reputation in a large webmail service. Proc. of the 3rd Conference on Email and Anti- Spam (CEAS), 2006.

[30] S. Hershkop and S. J. Stolfo. Combining email models for false positive reduction. Proc. of the 11th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), pages 98– 107, 2005.

[31] T. R. Lynam and G. V. Cormack. On-line spam filter fusion. Proc. of the 29th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), pages 123–130, 2006.

[32] R. Segal. Combining global and personal anti-spam filtering. Proc. of the 4th Conference on Email and Anti-Spam (CEAS), 2007.

[33] H.-Y. Lam and D.-Y. Yeung. A learning approach to spam detection based on social networks. Proc. Of the 4th Conference on Email and Anti-Spam (CEAS), 2007.

[34] L. H. Gomes, R. B. Almeida, L. M. A. Bettencourt, V. Almeida, and J. M. Almeida. Comparative graph theoretical characterization of networks of spam and legitimate email. Proc. of the 2nd Conference on Email and Anti-Spam (CEAS), 2005.

[35] J. Golbeck and J. Hendler. Reputation network analysis for email filtering. Proc. of the 1st Conference on Email and Anti-Spam (CEAS), 2004.


Models for Integrating Social Networking in Higher Education

Authors

Andreas Veglis Media Informatics Lab, School of Journalism & MC, Aristotle University of Thessaloniki

[email protected] Thessaloniki, 54006, Greece

Abstract

Today Information and Communications Technologies have been adopted in every level of education. Social networking services are one of the most widely used Web 2. Services. The aim of this paper is to study the use of social networking in higher education. Specifically it will investigate the advantages and the disadvantages of employing social networking in higher education. Furthermore the paper discusses methods for integrating social networking in higher education. The most prominent social networking services, namely, Facebook and Twitter are presented and discussed.

Key Words

Social networking, higher education, Facebook, twitter.

I. INTRODUCTIONOver the last 30 years with the introduction of personal computers, internet and its services,

Information and Communications Technologies (ICTs) have been adopted at every level of education. The ICTs are the modern digital technologies for encoding, processing, storing, retrieving and transmitting information in digital form. It is widely recognized that ICTs offer tools able to mark important developments in the educational process, such as the transformation of the student from a passive receiver into an active participant in the educational knowledge creation, a feature more suitable for the future citizens of the information society [1], [2].



Education has always been associated with the use of technology. As new tools, services and devices (Smartphones, tablets, etc) are adopted by users, educators have more options at their disposal for creating innovative practices in education [3]. ICTs can be employed to facilitate various educational practices, best described by the term e-learning. The term e-learning includes all forms of electronically supported learning and teaching. The information and communication systems serve as specific media to implement the learning process [4]. E-learning covers both out-of-classroom and in-classroom educational experiences via ICTs. It includes distance learning as well as hybrid or blended learning [3]. It can be self-paced or instructor-led and includes media in the form of text, image, animation, streaming video and audio.

An important part of ICT applications and tools is based on the World Wide Web. The first generation of such tools, included static websites, broadcast video, email, forum discussions, etc., and provided some sort of communication between users, but did not support effective interaction and collaboration. Users were characterized as passive consumers of content and the tools that were employed have been described with the term Web 1.0 [5]. But during the last decade the technology of the web and the internet in general, experienced significant transformation with the emergence and adoption of Web 2.0 applications. The technological changes that included Web 2.0 resulted in changes in communication and learning.

Closely associated with Web 2.0 is the term E-Learning 2.0, which Downes defined it as an interlocking set of open source applications, where learning is becoming a creative activity [6]. This is accomplished when users employ Web 2.0 tools and services in collaborative learning activities for autonomously producing learning content and use it for their own learning objectives [7]. Web 2.0 tools were quickly adopted and used in the higher education. Blogs, Wikis, social bookmarking and RSS where the first choices. Eventually these tools were implemented in the Learning Management Systems (LMSs) that are widely used in all levels of education [1]. Other Web 2.0 tools, like collaborative authoring tools, multimedia sharing services and social networking are being used as external services (for example Google Docs, YouTube, and Facebook).

One of the most widely used Web 2.0 tools is social networking [8]. Higher education like any other section of education has been influenced by the fast growing wave of social networking. This can be easily related to the fact that students belong to the group that exhibits the highest percentages of social networking usage [9]. In most cases universities are using social networking to disseminate information concerning their programmes and thus attracting potential students. On the other hand the actual introduction of social networking in the educational process is not considered to be an easy task [10], [11].

The aim of this paper is to study the use of social networking in the higher education. Specifically the paper proposes and discusses methods of integrating social networking in higher education. Special attention is given to Facebook and Twitter. The rest of the paper is organized as follows. Section II presents the issue of social networking. The advantages and disadvantages of integrating social networking in the learning process are examined in the following section. Section IV includes models for integrating social networking in higher education. Future extensions of this work as well as conclusions are included in the last section of the paper.

Integrating Social Networking in higher Education Andreas Veglis


II. SOCIAL NETWORKING The term Web 2.0 was proposed in 2004 and aimed in separating modern technologies (such as

wikis, blogs, sharing multimedia files, RSS feeds, etc.) that support greater interactivity, flexibility, collaboration and participation from the previous generation of Web 1.0 [2]. The term describes the tendency of change in the use of technology and the WWW in web design that aims to enhance creativity, communications, secure information sharing, collaboration and functionality of the WWW. Web 2.0 tools can be organized into the following categories: Blogs, Wikis, social bookmarking, RSS and other information technologies, collaborative authoring tools, multimedia file sharing services, social networking and social presence, bricolage and applications mashups [7], [12], [13]. Researchers suggested that Web 2.0 tools have penetrated people’s private and professional lives, but have also started transforming learning patterns [14]. Social networking is considered to be one of the basic categories of Web 2.0.

Social networking services can be defined as internet-or mobile device-based social spaces, designed to support communication, collaboration and content sharing across users [14]. They are web sites that facilitate the building of social networks or social relations among internet users that share similar interests, activities, backgrounds, or real-life connections (http://en.wikipedia.org/wiki/Social_ networking_service). Users are allowed to construct public or semi-public profiles within a bounded system, articulate a list of other users with whom they share a connection, and view and traverse their list of connections and those made by others within the system [8]. Prominent examples of social networking services include Facebook, Google+ and MySpace (for social Networking and socializing), LinkedIn (for professional networking), Academia and ResearchGate (for academic networking), and Elgg (for knowledge accretion and learning). The most well known and employed social network is Facebook, which in September of 2012 reached 1 billion registered users [15].

Researchers have studied many aspects of Facebook. Facebook research may be organized into four categories: social networking and social capital, identity construction, concerns with privacy and the potential use of Facebook for academic purposes [10], [16]. It is worth noting that the latter category includes the use of Facebook in university libraries.

Although it appeared later than Facebook, Twitter is another example of social networking service that became quickly very popular among users [17]. Twitter is a social networking and micro-blogging service that enables its users to send and read other users' updates, known as tweets. Micro-blogging is characterized by three features, namely, information sharing, information seeking, and friend-ship wide relationships [18]. Twitter is often described as the "SMS of Internet", in that the site provides the back-end functionality to other desktop and web-based applications to send and receive short text messages, often obscuring the actual website itself. Tweets are text-based posts of up to 140 characters in length. Updates are displayed on the user's profile page and delivered to other users who have signed up to receive them. Users can send and receive updates via the Twitter website, SMS, RSS (receive only), or through applications. The service is free to use over the web, but using SMS may incur phone services provider fees [19].

While there has been an increase in the usage of Twitter at higher education worldwide, a report from Faculty Focus [20] indicated that Twitter‘s potential pedagogical uses has yet to be exploited, since most institutions are currently using it for sharing information among students



and as a real time news source. Several researchers have studied the use of Twitter, but the

III. ADVANTAGES AND DISADVANTAGES OF USING SOCIAL NETWORKING IN THE EDUCATIONAL PROCESS

Although social networking services appear to be very appealing for integrating them in the educational process the tendency is to employ them as a means of information dissemination and not to actually encapsulate them in the educational process [21]. For many years universities have employed Learning Management Systems (LMSs) in order to support distance but also hybrid learning. LMSs are content management systems especially tailored for the educational needs. An LMS is the infrastructure that delivers and manages instructional content. It also identifies and grants access to registered users, tracks their progress, collects and presents data for supervising the learning process of an educational organization [22]. Thus universities are trying to exploit social networking capabilities while they continue to use LMSs. And this move is quite natural since LMSs are what the members of the educational process have been using for the past 20 years. They know their strengths and their drawbacks.

Many universities around the world have profiles in Facebook in order to promote their activities. Sharing information for various events seems to be the main use of Facebook by universities [10]. Since the majority of university students are using Facebook, it is quite natural to use this channel of communication in order to disseminate news and information. In the Department of Journalism and Mass Media Communication in Aristotle University of Thessaloniki, all students are members of the departments’ group in Facebook (http://www.facebook.com/#!/groups/ 43284533114/?fref=ts). All department’s announcements are also published in the wall of this group. The majority of the department’s teaching staff is a member of this group and communicate directly with the students via the wall of this group.

Although it seems a good idea to move educational content in to Facebook, it is not certain that students will appreciate this action. Many students appear to prefer keeping the educational process separately from their online social activities. Stutzman in his research reported a critical finding regarding the desire of students to treat academic content as a separate from their social interactions [23].

Also Facebook is perceived a shared-space (outside of the university’s controlled online systems) and students tend to feel more relaxed in their communication with other students as well as with the instructors [10]. This may lead to open discussions and thus valuable interactions between learners and instructors. But it may also have some negative aspects. Specifically students have many distractions in the Facebook environment, which may cause them to lose concentration more easily. Also the university has no control on the Facebook and this may cause problems as far as the availability of the online system is concerned and other software compatibility problems that students might face with their personal computers [10].

One other issue that must be mentioned is that in the neutral environment of Facebook, the boundaries between learners and instructors are blurred, and learners may take initiatives in contributing to the course content or explaining course material to other fellow learners. Of course such actions are welcomed in the learning process, since they lead to deeper understanding of the course material. But students must be ready to take such responsibilities and be able to behave in a suitable manner.



IV. INTEGRATION MODELS New technologies may fall into two categories. They either substitute an existing technology

that is already in use, or they add new capabilities and thus extent an existing technology [24]. In this context this section will attempt to outline the possible solutions for the integration of social networking in higher education.

If we try to describe the current situation as far as the integration of social networking in higher education is concerned, we can say that social networking services are currently employed in parallel with existing LMS infrastructure that supports distance and hybrid education. Thus a certain group of users is part of a learning community and also employs Facebook or other social networking services for communication purposes (see figure 1). It is worth noting that various plugins that allow the flow of information between LMSs (Blackboard, Moodle, etc) and Facebook are now available, but they offer limited connectivity between the two platforms. Although this may look like a logical development given the popularity of Facebook and its proven communication capabilities, it cannot be considered as an ideal solution. We must not forget that the use of Facebook at his current form offers many advantages but also has many disadvantages (as described in the previous section).

FIGURE 1: USE OF SOCIAL NETWORKING SERVICES IN CONJUNCTION WITH EXISTING LMS. One would expect that the next logical evolution is to either integrate education in the social

networking services, or integrate social networking in education. The problem is that social networking services are operated by business corporations that aim in profiting by the development of successful social networking services. Thus it would be difficult for society to accept Facebook as a potential education institute. One must not forget that Facebook includes many advertisements in its environment, which is its main source of revenue. The described scheme is depicted in figure 2. This option requires the introduction of additional features in social networking services that will support, course content organization and presentation as well as evaluation and monitor of the learning process.



FIGURE 2: INTEGRATION OF A LEARNING PROCESS IN AN EXISTING SOCIAL NETWORK. The second option is for universities to develop their own social networking service. Although

this solution seems to have many advantages since it adopts the communication features of social networking in the existing educational context, it does not appear to have many possibilities of success. It is not easy to convince students to start using a new social networking service, when they are already using one (Facebook most probably). And if they are forced to adopt it, the most probable outcome is that they will eventually abandon it. In the previous years we have all witnessed various attempts from Google to introduce a successful social networking service. Their latest attempt was Google+, which although it has initially grabbed public attention it did not succeed in becoming a major force in social networking [25].

Thus the idea is not to reinvent social networking, but to try to combine the existing established status quo with the learning process. For example one proposal could be to introduce sub social networks in an existing social network that will be connected with a university (see figure 3). A social networking site can create a sub network for each university that is interested in participating in such a scheme. The social networking service will offer to its users additional features related to education and will also attract new users they were not registered users of the social networking service. One the other hand the university will be able to use all the social features offered by the social networking service. Finally the learner/user will be using a familiar environment, with all the friends and connections that he had made before entering the university. He will also be able to participate to more than one educational sub networks, depending on its current studies (for example studding for a BA degree, while also attending a part-time post graduate programme offered by a different institution). Of course special attention must be given in order for the user to have a clear distinction between the educational content and communication and the rest of the social traffic that originates from the social networking service. The learning material can be sustained in the university’s IT infrastructure and integrated in social networking pages as mashup application or accommodated in the social networking sites’s IT infrastructure.



FIGURE 3: CREATION OF EDUCATIONAL SUB-NETWORK IN AN EXISTING SOCIAL NETWORKING SERVICE. As far as twitter in concerned things are quite different. Since Twitter possesses specific

(limited in comparison with other social networking services) characteristics it can be employed as a supplemental communication tool in the educational process. Already they are available plugins for major LMSs (for example Blackboard, Moodle, etc) that allow the flow of information from Twitter to specific modules of the learning platforms.

V. CONCLUSION AND FUTURE RESEARCH DIRECTIONS It is obvious that the integration of social networking in the educational process is a big

challenge for higher education institutes. There are a lot of benefits that may arise from this integration. The advantages offered by the use of social networking services can lead to a more learner centered model in the higher education, in which the learning process will adapt to the learner’s needs [49]. Cohen believes that LMSs must be enriched in order to support social networking and collaboration tools [24]. This is already happening, with major LMSs (for example, Blackboard, Moodle, etc) gradually adopting various tools and services that are offered by social networking services. But every change must occur gradually in order for the learners and the instructors to have time to learn and apply the new added features in the learning process. Future research directions of this work may include the implementation of online courses entirely in a social networking site, for example Facebook. Thus we will be able to assess the impact that will have in the learning process and also evaluate learners’ and instructors’ reactions. The results from this study will be compared with results from the evaluation of the same course offered by a standard LMS. Also surveys must be conducted among university students in order to evaluate their attitude towards the use of Facebook in higher education. The generation that has grown up using social networking is now starting to enter universities thus their attitude may be different from students that were introduced to social networking as adults. Based on the findings of the previous mentioned studies we will be able to reevaluate the



way of integrating social networking in higher education.

This paper has investigated the integration of social networking in higher education. It is well accepted that social networking services constitute a considerable force in today’s communication world. And without a doubt Facebook with its 1 billion users is social networking most important player. The learning process can significantly benefit from the incorporation of tools and services that support communication between the participating parties. The shift from a teacher centric system to a student centric system is facilitated by the use of information and communication technologies. The introduction of social networking has changed considerably the technology world and the way we interact with each other online. As a result existing technology that has been employed for a considerable time period in the learning process needs to be redefined and enriched with new features [24]. In this context various solutions where explored in order for universities to be able to take advantage of the communication capabilities offered by social networking services. These solutions will without doubt include existing technological infrastructure (that is LMS) already in use by universities today. LMS must support the creation, sharing, enrichment, and commenting of user content. Learners must be able to post questions, documents, best practices and locate useful information. These features will enrich existing LMSs and make them more social oriented [24]. The most promising solution involves the creation of educational sub-networks within existing social networking sites that will facilitate learning in conjunction with LMS. As far as micro-blogging services, like Twitter is concerned the study acknowledges its potentials and proposes its use in conjunction with existing LMSs, as a supplemental communication tool for the learning process. All the above can be accomplished by a close collaboration between social networking services, LIM vendors and universities.

REFERENCES [1] Hirtz, S. (2008). Education for a Digital World: Advice, Guidelines, and Effective Practice from Around

the Globe. Commonwealth of Learning.

[2] Veglis, A. (2012). Using Web 2.0 tools in the Department of Journalism & Mass Media Communication - Educational practices and perspectives. presented in the Scientific Workshop on "Undergraduate education and educational practices in Aristotle University of Thessaloniki (in Greek), April.

[3] Delich, P., Kelly, K., & Mclntosh, D. (2008). Emerging Technologies in E-learning, Education for a Digital World. Commonwealth of Learning.

[4] Tavangarian D., Leypold M., Nölting K., & Röser M. (2004). Is e-learning the Solution for Individual Learning? Journal of e-learning.

[5] Usluel, Y.K., & Mazman, S.G. (2009). Adoption of Web 2.0 tools in distance education. Procedia Social and Behavioral Sciences 1, 818-823.

[6] Downes, S. (2005). E-Learning 2.0. eLearn Magazine, October. Retrieved from http://elearnmag.acm.org/featured.cfm?aid=1104968.

[7] Blees, I., & Rittberger, M. (2009). Web 2.0 Learning Environment: Concept, Implementation, Evaluation. eLearning Papers. No 15, June.

[8] Boyd, D.M., & Ellison, N.B. (2008). Social Network Sites: Definition, History, and Scholarship. Journal



of Computer-Mediated Communication, Volume: 13, Issue: 1, 210-230.

[9] Observatory for digital Greece (2011). Internet use in Greece. Report.

[10] Bosch, T. (2009). Using online social networking for teaching and learning: Facebook use at the University of Cape Town. Communicatio: South African Journal for Communication Theory and Research, 35:2, 185-200.

[11] Franklin, T., & Van Harmelen, M. (2007). Web 2.0 for Content for Learning and Teaching in Higher Education, Fanklin Consulting and Mark van Hermelen

[12] Anderson P. (2007). What is Web 2.0? Ideas, technologies and implications for education. JISC Technology and Standards Watch. February.

[13] Bartolome, A. (2008). Web 2.0 and New Learning Paradigms. eLearning Papers No 8, June.

[14] Redecker, C., Alla-Mutka, K., Bacigalupo, M., Ferrari, A., & Punie, Y. (2009). Learning 2.0 The Impact of Web 2.0 Innovations on Education and Training in Europe. Final Report, Institute for Prospective Technological Studies. Retrieved from http://ftp.jrc.es/EURdoc/JRC55629.pdf

[15] Vance, A. (2012). Facebook: The Making of 1 Billion Users. Bloomberg Business week, October 04. Retrieved from http://www.businessweek.com/articles/2012-10-04/facebook-the-making-of-1-billion-users

[16] Susilo, A. (2008). Use of Facebook for Academic network learning in Universitas Terbuka-Indonesia. AAOU Journal, vol. 3, no 2, September, 99-114. Retrieved from http://lppm.ut.ac.id/ htmpublikasi/03aauo32.pdf

[17] An, J., Cha, M., Gummadi, K., and Crowcroft, J. (2011). Media landscape in Twitter : A world of new conventions and political diversity. Artificial Intelligence, Volume: 6, Issue 1, 18-25.

[18] Saeed, N. & Sinnappan, S. (2011). Adoption of Twitter in higher education – a pilot study. In G. Williams, P. Statham, N. Brown & B. Cleland (Eds.), Changing Demands, Changing Directions. Proceedings ascilite Hobart 2011. (pp.1115-1120).

[19] Veglis, A. (2012). Journalism and Cross Media Publishing: The case of Greece. In E. Siapera & A. Veglis (Ed.), The Wiley-Blackwell Handbook of Online Journalism (pp. 209-230). Blackwell Publishing.

[20] Faculty-Focus. (2010). Twitter in higher education 2010: Usage habits and trends of today‘s college faculty. Annual Survey on the Popular Microblogging Technology. Retrieved from http://www.facultyfocus.com/free-reports/twitter-in-higher-education-2010-usage-habits-and-trends-of-todays-college-faculty/

[21] Thompson, J. (2007). Is education 1.0 ready for web 2.0 students? Innovate Journal of Online Education, Vol. 3, No. 4.

[22] Watson, W.R., Watson, S.L., (2007). An Argument for Clarity: What are Learning Management Systems, What are They Not, and What Should They Become? TechTrends, March/April, Vol. 51, No 2, pp. 28-34.

[23] Stutzman, F. (2008), The vibrancy of online social space, In B. Rigby (Ed.), Mobilizing generation 2.0: a practical guide to using web 2.0 technologies to recruit, engage & activate youth. New York, NY: Jossey-Bass.



[24] Cohen, E. (2010). Is the LMS Dead?. Chief Learning Officer, October.

[25] Johnson, B. (2012). This why Google is losing the future. GIGAOM, March, 16th. Retrieved from http://gigaom.com/2012/03/16/this-is-why-google-is-losing-the-future/


Andreas A. Veglis was born in Thessaloniki, Greece in 1964. He received his BSc in Physics (1988), MSc in Electronics and Communications (1992), and PhD in Computer Science (1995), all from Aristotle University. He is a Professor, head of the Media Informatics Lab in the Department of Journalism & Mass Media Communication at the Aristotle University of Thessaloniki in Greece. He is also head of the postgraduate programme and Deputy Chairman of the Department. His research interests include information technology in journalism, new media, course

support environments, and distance learning. Professor Veglis is the author or co-author of ten books, he has published 50 papers on scientific journals and he has presented 65 papers in international and national Conferences. He is one of the co-editors in the Handbook of Global Online Journalism (Wiley-Blackwell). He has been involved in 11 national and international research projects.


Wireless Sensor System According to the Concept of IoT -Internet of Things-

Authors

Juan Felipe Corso Arias Mechatronics/Engineering, Military Nueva Granada University

[email protected] Bogotá, Colombia

Yeison Julian Camargo Barajas Telecommunications/Engineering, Military Nueva Granada University


Juan Leonardo Ramirez Lopez Telecommunications/Engineering, Military Nueva Granada University


Abstract

This article presents the design of a wireless communication system, responding to the sensor concept applied to a scaled industrial process where temperature variables were used. The sensors are connected to the internet (IoT) to be monitored remotely from anywhere in the world. The sensor data is downloaded from the cloud using a graphical programming platform to control and communicate the system with a programmable logic controller (PLC), which performs the actions according to the temperature value (set point) of the sensors. The monitoring process was performed with a SCADA system and the modeling of the communication system was performed using the formalism of Petri nets, as a system that responds in terms of discrete events.

Key Words

IoT, Petri, PLC, SWI.



www.elvedit.com 328

I. INTRODUCTION The trend of the market is the information to be available independently of the place or the

geographic location. For this reason, currently, the internet is used to bring a real time interaction among devices that will not be possible with other mediums [1]. Simply, all the information gathered from the sensors must be available in the cloud to be managed and controlled. Thus, a central point of management exists where the information from the sensors remotely distributed is stored. Furthermore, the industrial processes make necessary to implement wireless communication systems (due to the hostile environment and the difficult access to the places) to transmit the signals generated by the sensors making up the control loop [2]–[8]. Hence, the modeling, design and implementation of a remote wireless system applied to an industrial process is done. Then, the reliability of the system can be analyzed. A SCADA system supervises the process and a PLC [9] executes the event generated in the control. The modeling of the communication was performed using Petri nets [10]–[14]. It allowed analyzing the viability and reliability of the system. The rest of the paper is organized as follows: Section II presents a description of the implemented system according to the IoT, the mathematical modeling using Petri nets, the simulation of the proposed Petri net, a networking analysis and a description of the developed control system. Section III shows the analysis of the results obtained. Finally, Section IV shows the conclusions.

II. DEVELOPMENT In this section we will describe the implemented sensor’s system [15]. The system uses three

type of sensors (Thermocouple, thermistor and Integrated Circuit LM35) for temperature sensing. The sensors send the information to the Internet, more exactly to the Xively’s servers (a platform designed for the IoT). Thus, the information gathered can be monitored from anywhere in the world. Once the data is stored in the internet a computer is used to download it and process it to be used in a temperature control system. The project goal is to develop a redundant system of temperature measurement. Hence, if a failure exists in one of the sensors the system is able to continue online with the rest of the sensors (using the average of the temperature gathered). The average of the temperature captured from the sensors is used as a threshold value or set point that enables the system to take a control decision. Then, the processed information is sent to the PLC to perform an action. The implemented action in the system is to open or close a cylinder according to the temperature value. Furthermore, the system implements a SCADA system that allows plotting and visualizing the information available from the internet corresponding to the temperature values gathered by the sensors. It also allows controlling the system manually, then the administrator is able to change the action whenever he wants. Figure I shows a flow chart and the topology of the implemented systems.

Wireless Sensor System According to the Concept of Internet of Things Juan Felipe Corso, Yeison Julian Camargo and Leonardo Ramirez Lopez


a) System’s Flow Chart b) Real Topology

FIGURE I: IMPLEMENTED SYSTEM.

A. System Modeling Using Petri Nets of Type (P/T) SWI (Sensors Wireless Industry). A petry net (푃/푇), is a tuple 푁 =< 푃,푇,푃푟푒,푃표푠푡 > where:

푃 is a set of finite sites. 푇 is a set of finite transactions disjoined of 푃 such that 푃 ∩ 푇 =

푃푟푒,푃표푠푡 ∈ 푁| |∗| | are incidence matrixes of 푁. 퐶 = 푃표푠푡 − 푃푟푒, where 퐶is defined as the incident matrix of 푁. There exists an arc with weight 푛 > 0 from a place 푝 ∈ 푃, to a transition 푡 ∊ 푇 if and only if 푃푟푒[푝, 푡] = 푛푤푖푡ℎ푛 > 0, and there exists an arc with weight 푛 > 0 from a transition 푡 ∊ 푇 to a place 푝 ∈ 푃 if and only if 푃표푠푡[푝, 푡] =

푛푤푖푡ℎ푛 > 0. Thus, the set of arcs of 푁is defined as in equation 1. 퐹 ∶= {(푝, 푡) ∈ 푃.푇 ∕ 푃푟푒[푝, 푡]} > 0} ∪ {(푡,푝) ∈ 푇.푃/푃표푠푡[푝, 푡] > 0} (1)

Control Supervisor

Access Point

PLC

Xively (Internet)

Router Router

Capture System and

Data

Wireless Communication

Shield

Sensor 2

Sensor 3

Sensor 1

Instrumentation Variables

(Temperature)

Temperature 1 (Thermocouple) Temperature 2 (Thermistor) Temperature 3 (LM35)

SCADA System

Wi-Fi (802.11g)

LAN NETWORK

WAN NETWORK

Control and Supervision

PLC

Data Capture with Wireless Shield

Router



www.elvedit.com 330

This allows an alternative definition of 푁 that fits better for the graphical representation of the Petri Net. The modeling of the communication system was done using the WoPeD software [16]. It is a free software under LGPL license. It is aimed to provide a tool for the modeling and analysis of the control flow of processes and the description of the resources by using control flow networks (an extension of the Petri nets). The Table I shows the nomenclature used to make the equations clear.

TABLE I: NOMENCLATURE

Name Acronym Place p Transition t1,2,3,4

Subprocesses(Wi-Fi,Ethernet) sub1,2 Source p1 Thermocouple p2 Thermistor p3 LM35 p4 Data Capture with wireless shield p5 Router p6 Xively p7 PC with Labview p8 PLC Siemens p9

The model is graphically depicted in Figure II; the (푝 ) token represents the initial marking of

the system. The transitions (푡 , , , ) represent the events of changing from one place to another with a default weight in the arcs not labeled defined by푤(푝 , 푡 ) = 1. The subset (푠푢푏1) represents the wireless communication between the system for data capture with a wireless shield and the layer 3 device or router. The communication used the standard IEEE 802.11g. The subset (푠푢푏2) represents the medium access protocol in the Ethernet architecture (CSMA/CD) for the LAN network where the control process is performed. In this case it is necessary a LAN network due to the PLC does not work with a wireless module. Based on a Petri net with a type (푃/푇), the model shown in the Figure II is developed for the System of Wireless Industrial sensors (SWI) [17]. The places and the transitions are defined according to the equation 2 and equation 3.

푝 = {푠푒푛푠표푟푠(푡ℎ푒푟푚표푐표푢푝푙푒, 푡ℎ푒푟푚푖푠푡표푟, 퐿푀35);퐷푎푡푎푐푎푝푡푢푟푒푤푖푡ℎ푤푖푟푒푙푒푠푠푠ℎ푖푒푙푑;푅표푢푡푒푟;푋푖푣푒푙푦;

푃퐶푤푖푡ℎ퐿푎푏푣푖푒푤;푃퐿퐶푆푖푒푚푒푛푠} (2)

푡 = {푡 ,푡 ,푆푢푏 ,푡 ,푆푢푏 ,푡 } (3)



c) sub1 d) sub2

FIGURE II: GENERAL PETRI NET

Consequently, the petri net showed in the Figure II is described mathematically as follows: 푊 ∶ 퐹 → 푁 ∖ {0} Where W is a defined weight function. 퐹 = {(푝 , 푡 ), (푡 ,푝 ), (푡 ,푝 ), (푡 ,푝 ), (푝 , 푡 ), (푝 , 푡 ), (푝 , 푡 ), (푡 ,푝 ), (푝 , 푆푢푏 ), (푠푢푏 ,푝 ), (푝 , 푡 ), (푡 ,푝 ), (푝 , 푠푢푏 ), (푠푢푏 ,푝 ), (푝 , 푡 ), (푡 ,푝 )} → 퐴푟푐푠 푊푒푖푔ℎ푡 = 푊(푝 , 푡 ) = 1,푊(푡 ,푝 ) = 1, … . .푊(푡 ,푝 ) = 1 푃푟푒,푃표푠푡 ∈ 푁| |∗| | are incidence matrixes of N.

The total communication of the system is represented in a matrix form. In the matrix the columns represents the transitions, the rows represents the places and the cells the connection between them. The previous incidence matrix (shown in Table II) is the result of the arcs built from the places (푝) to the transitions (푡) , i.e., 푝푡. Initially, there exists a place 푝 that represents the environment or the universe where the Petri net is located and through an arc directed to the transition푡 , the cycle is started. In this case the (푡 ) transition has an input place (or preconditioned) such that when it is activated the sensors are ready to send the information to the microcontroller. The next cycle represents the send of data from the three sensors (푝 ,푝 ,푝 ,) to the system of capture and data processing (푝 ) with wireless shield, symbolically we have the following representation: 푝 ,푝 ,푝 ,(푡 ) (input to 푡 ). In the matrix the input to the transition (푡 ) from the sensors (푝 ,푝 ,푝 ,)is shown which their directed arcs have a weigh of 1. The cycle is continuous until the signals from the sensor are received by the PLC, who it is in charge of executing the final action, i.e., activate a cylinder according to the temperature or set point.



www.elvedit.com 332

TABLE II: PREVIOUS INCIDENCE MATRIX

Pre t1 t2 sub1 t3 sub2 t4 p1 1 0 0 0 0 0 p2 0 1 0 0 0 0 p3 0 1 0 0 0 0 p4 0 1 0 0 0 0 p5 0 0 1 0 0 0 p6 0 0 0 1 0 0 p7 0 0 0 0 1 0 p8 0 0 0 0 0 1 p9 0 0 0 0 0 0

Table III shows the posterior incidence matrix columns. In this matrix the columns represents

the transitions, the rows represent the places and the cells represents the connection between them. The “posterior incidence matrix” is built from the arcs formed from the (t) transitions to the places (푝), such that 푡푝. Initially, there exists a transition 푡 that represtns the event of changing from the starting place 푝 to the places 푝 ,푝 ,푝 (푡ℎ푒푟푚표푐표푢푝푙푒, 푡ℎ푒푟푚푖푠푡표푟, 퐿푀35). In this case, the transition (푡 ) has three output arcs (or post conditionals) such that when it is activated the sensors are ready to send the information to the microcontroller 푡 푝 ,푝 ,푝 . The next cycle represents the send of data from the three sensors (푝 ,푝 ,푝 ,) to the capture and data processing (푝 ) with wireless shield; symbolically we have: 푡 푝 (output from 푡 ). In the matrix, the output from the (푡 ) transition to the sensors (푝 ,푝 ,푝 ,)is shown such that their arcs have a weight of 1. The cycle is continuous until the entire process of the system is ended. This is the control part that depends on the signals from the sensors.

TABLE III: POSTERIOR INCIDENCE MATRIX

Post t1 t2 sub1 t3 sub2 t4 p1 0 0 0 0 0 0 p2 1 0 0 0 0 0 p3 1 0 0 0 0 0 p4 1 0 0 0 0 0 p5 0 1 0 0 0 0 p6 0 0 1 0 0 0 p7 0 0 0 1 0 0 p8 0 0 0 0 1 0 p9 0 0 0 0 0 1



The incidence matrix is formed by the subtraction of the post – pre matrixes. Consequently, the

pair of the matrix 푝표푠푡(푝1, 푡1 = 0) is subtracted from the pair of the matrix 푝푟푒(푝1, 푡1 = 1); ⟹0 − 1 = −1 and the same procedure is executed until completing the incidence matrix as shown in Table IV. The incidence matrix shows the entire network to be unidirectional, then, 푝1 has a direction towards푡1, but not on the opposite direction (-1).

TABLE IV: INCIDENCE MATRIX C: = POST - PRE

C t1 t2 sub1 t3 sub2 t4 p1 -1 0 0 0 0 0 p2 1 -1 0 0 0 0 p3 1 -1 0 0 0 0 p4 1 -1 0 0 0 0 p5 0 1 -1 0 0 0 p6 0 0 1 -1 0 0 p7 0 0 0 1 -1 0 p8 0 0 0 0 1 -1 p9 0 0 0 0 0 1

Furthermore, the incidence matrix allow the places and transition to know which are the arcs

that they are connected to and to obtain information from them to determine which are the places for input and output in a transition. Finally, it is convenient to define the tokens as vectors which inputs are integer numbers, assuming that the places form a totally ordered set. In this way, the initial token in (푃/T) IWN for the vector 푀 is defined by (1,0,0,0,0,0,0,0,0) and the cycle continues until completing the occurrences sequence shown in Table V. In other words, when an enabled transition is triggered, it will change the distribution of the signals (tokens). In this way, a trigger’s sequence will produce a marking sequence and consequently the Table V is formed:

TABLE V: SEQUENCE OF OCURRENCES (MO)

1

t1 ->

0

t2 ->

0

s1 ->

0

t3 ->

0

s2 ->

0

t4 ->

0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1

Conventionally, the techniques for the analysis of Petri models are classified in: I) numeration;

II) transformation; III) structural analysis and IV) simulation. For the Petri net (P/T) IWN, the



www.elvedit.com 334

enumeration method was used. It is based on a graph called “coverability graph”. It represents each marking vector in the network and the transitions among them as shown in the Figure III. One method to verify that the petri net has the desirable properties is to build a coverability graph. The graph consists of nodes made of binary sequences (0s and 1s) connected by the transitions defined in the network. The series start from the node, which elements are zeros and they represent all the possible states that can be obtained when the network is working under all the desirable conditions. The sequence of occurrences shown in Table V is one the possible combinations in the coverability graph.

FIGURE III: COVERABILITY GRAPH

B. Processing and Data Capture System with Wi-Fi

The capture and data processing system was developed with the C++ programming language, including the libraries for the communication protocols, such as: DHCP and HTTP that belong to the application layer in the OSI model. TCP was used in the transport layer. The wireless communication between the sensors and the router is established using the Wi-Fi libraries (slightly modified) of the capture and data processing system to identify the SSID and the password for the WPA or WEP security protocol that is being connected to the Wi-Fi shield. Finally, the programming necessary to send all the parameters to the cloud was done. The process of sending the data to the Internet has different phases. Initially the temperature is captured by the thermocouple and passed to the (Integrated Circuit) IC MAX 6675 which performs the cold junction compensation (it compensates the dependence with the environmental temperature inherent to the measure), amplifies and converts to digital the temperature obtained from the thermocouple. The MAX IC passes the data to the microcontroller ATMEGA 328P through the following serial port: SPI of 12 bits and 0.25 centigrade grades of resolution. Lastly, the microcontroller passes the data to the Wi-Fi wireless shield that sends the information to a wireless router.



C. Analysis of Dataflow

Pathchar [18] is a Linux utility similar to traceroute but focused in the network performance measurements. The output of the Pathchar software is shown in Table VI. This output shows the measurements of the possible paths that the data can take when traveling from a sensor to the cloud servers. The information is analyzed with all the metrics among the hops involved from the source of the packets (sensor) to the destination of the packets (IoT web portal - Xively). The tool sends 45 different packets sizes in the range from 64 to 1500 bytes (1500 is the MTU in the local host). The software uses 32 different sets of this packets per hop. Thus, 11.520 packets are sent and the same number of answered are shown by the software.

TABLE VI. TRAFFIC MEASUREMENTS Link Host IP BW Latency Drodped Queuing

0 System of processing and capture of Data

Wi-Fi

192.168.0.3 39 Mb/s 124 µs

1 Modem Wi-Fi 192.168.0.1 722 Kb/s 14.8 ms 11 % 2 10.32.0.26 3 *201.244.1.150 *10.5.4.70 *201.244.1.5 490µs

4 sta.etb.net.co 10.5.4.74 27ms 5 Edge3.Miami1 4.59.82.113 *4.69.138.77 73 µs 42%

6 Edge2.Miami2 4.69.138.109 571Mb/s 345 µs 1.23 ms 7 globalcrossing 4.68.111.122 178 Mb/s 156 µs 8 ae9.scr4.gblx.net 67.16.147.129 30.1 ms 9 po2.gblx.net 67.17.95.214 161 ms 10 INTERNAP 64.215.30.78 *216.52.255.46 49 Mb/s 4 µs 1.55ms

11 pnap.net 216.52.255.110 135 ms 12 Logmein.net (xively) 63.251.195.114 * 210.52.233.121

In this part of the paper we present an analysis carried out to find the performance of the

system. A software was used to find the RTT (Round Trip Time) and the lost packet rate in the communication from the sensors to the Internet, specifically the Xively’s servers. The Smokeping [19] program was used to perform the measure of RTT in the traffic from the sensors (source) to the Xively’s servers (destination) during 30 hours. Figure IV depicts the output of the Smokeping program and the Petri nets simulator explained in section III. It indicates the average percentage of lost packets to be 0.6% which is very low. The average RRT was 177.7 ms with a standard deviation of 6.1ms. These values are relative low taking in to account the possible distances from the sensor to the central management point. The Table VII shows the values above explained. The RRT is the time that a sent packet expends in going from its source to return after going to



www.elvedit.com 336

its destination [20]. This value is a key aspect in the HTTP protocol that is used to send the data from the sensors to the internet. According to [21] the RTT time is calculated as in Equation 3

푅푅푇 = 퐿푎푡푒푛푐푦 + ( ) + 퐿푎푡푒푛푐푦 (3)

Figure IV: XIVELY’S RTT OVER HTTP (DAY 1)

TABLE VII. COMMUNICATION MEASUREMENTS

Lost Packet Rate RTT (ms) Standart Deviation (ms)

0.60 % 177.7 6.1 The average RTT (177.7ms) is shown in Figure IV, i.e., the average RTT from the sensor to the

Xively sever in the Internet. The figure shows a stable communication with some peaks. The HTTP protocol was used to transfer the information. The HTTP (Hypertext transfer protocol) protocol is also the base for the communication in the World Wide Web. In order to measure the



end to end delay the Smokeping software was used. The Figure V shows a capture of the HTTP protocol in the application layer towards the Xively servers (api.xively.com) with the Wireshark software. The PUT method (in the HTTP protocol) was used to send (or upload) the temperature data gathered by the sensors (220 bytes) to the servers located in the cloud. The figure shows how the data is send within the http protocol, in this case the data is pointed out by the red circles in the image. The PUT method allows writing a file with an established connection to the server.

FIGURE V: HTTP PROTOCOL.

Figure VI, shows the sequence of the TCP protocol in the transport layer. The black line represents the source port and the red point represents the destination port in the communication, in this case the destination port in the Xively portal. The green lines represent the acknowledgement in the TCP process.

FIGURE VI: TCP SEQUENCE PROTOCOL



www.elvedit.com 338

D. The IoT

The main idea of the IoT is to integrate objects in a communication network [22]–[26]. In other words, it represents a new approach where all the things make part of the internet, working among them in real time. Technically, it is based on the integration of sensors, devices and household things that get connected to the internet through wired or wireless networks. Because of the world wide deployment of the internet the adoption of this technology is feasible and the cost of the implementation will be inexpensive. It will allow the sensors to be integrated in homes, workplaces, automation processes and so on. In this way, every object is able to be connected to a web environment (in this work the Xively’s webpage) to store all its information and show it in real time (if desirable). Figure VII shows the gathered data from the sensors directly to the internet. It allows filtering and plotting the data to find possible trends or critical situations. The data can be checked everywhere in the world independently the location of the sensors.

FIGURE VII: XIVELY´S WEB PAGE.

E. OPC and Real Time System in LabView

In this part of the work we aim to describe the system developed that connects the Internet data to the PLC through LabView (Figure VIIIa). The data from the cloud are downloaded with a PC with LabView to be processed and afterwards be sent to the PLC. The PLC is in charge of executing the temperature control programed with the LabView software through the OPC communication standard (communication between the PLC and LabView) [27]. The SCADA system is the GUI (Graphical User Interface) that allows the interaction with the sensors system and displays the gathered data. Hence, it controls and supervises the process. When the temperature value is above the threshold, the PLC performs the opening action of the system. Additionally, there exists new alternatives for SCADA systems that can be monitored from the cloud; one example is the webpage 3pdashboard. In this website it is possible to design the control system and plot the data gathered from the sensor that is stored in the Xively´s servers. The management and analysis of the data is executed in real time and within a webpage without the need of installing any software in a PC. The website has a graphical environment where it is



possible to create a control system for the sensor network. We also developed an application (in the 3pdashboard website) that plots the data gathered by our sensors. The application developed is shown in the Figure 8b.

a)

b)

FIGURE VIII: CONTROL OF THE SYSTEM.

III. RESULTS In this section we aim to describe the results obtained from the simulation of the Petri net. The

Workflow Petri Net Designer (WoPeD) software was used to simulate the proposed mathematical modeling of the Petri net corresponding to the implemented sensor’s system. The results of the simulation will allow to determine the robustness of the system. Based on the mathematical modeling of the Petri net (P/T) of type (P/T) SWI that corresponds to a workflow system, it was possible to carry out a dynamic analysis through the SWI simulation. In the simulation the RTT was used as continuous random variable. It showed the dynamic behavior of the petri net to follow a statistical pattern varying from a chaotic state in the first day (refer to Figure IX) to a stable state when augmenting the sampling periods (due to the higher number of events) and finally a Gaussian pattern in the 10TH day which keeps the same this point forward. It allows concluding that the data flow based on the RTT follows a robust behavior around the experimental median (λ) introduced in the simulation. The time between a pair of consecutives events has an exponential distribution with a (λ) parameter (in this case RRT) and independent



www.elvedit.com 340

in statistics terms. Figure VI shows the Gaussian pattern that indicates the stabilization of the communication system between the wireless sensors and the PLC.

a) Day 1 b) Day 5 c) Day 30

FIGURE IX: GAUSSIAN PATTERN.

IV. CONCLUSIONS In this work we presented a Wireless Sensor System that is able to be monitored all over the

world using the Internet. The system was also simulated with Petri nets. The results of the simulation described a pattern that goes from a chaotic state in the first day to a Gaussian state day after day (refer to Figure IX). The pattern denotes the stability and reliability of the industrial communication in environments using data networks. The results showed the system not to be suitable for industrial processes that require immediate responses times due to the delay found in the network based on the simulations performed (refer to Figure IV). But, according to the decrement of the delay the implementation of the system will allow using data networks in industrial processes with real time responses. The model could be implemented in conditions where the change in the measure (temperature, level, grades) respect to the time is not critical. Finally, the implemented system demonstrated that it was possible to build a control system (PLC) that centralizes the data gathered from sensors distributed around the world. Hence, the sensors can be monitored and controlled from anywhere in the world according to the concept of IoT.

V. ACKNOWLEDGMENT

This research was supported by Military Nueva Granada University. Specially by the Master Degree program from the Mechatronics Department

REFERENCES

[1] S. Cesar San Martin, F. Torres, R. Barrientos, and M. Sandoval, “Monitoreo y control de temperatura de un estanque de agua entre chile y España usando redes de alta velocidad,” REVISTA FACULTAD DE INGENIERÍA, UTA (CHILE), vol. 11, no. 1, pp. 41–46, 2003.



[2] L. Hou and N. Bergmann, “Novel industrial wireless sensor networks for machine condition monitoring and fault diagnosis,” Instrumentation and Measurement, IEEE Transactions on, vol. 61, no. 10, pp. 2787–2798, 2012.

[3] Z. Ke, L. Yang, X. Wang-hui, and S. Heejong, “The application of a wireless sensor network design based on zigbee in petrochemical industry field,” in Intelligent Networks and Intelligent Systems, 2008. ICINIS ’08. First International Conference on, 2008, pp. 284–287.

[4] G. Cena, A. Valenzano, and S. Vitturi, “Wireless extensions of wired industrial communications networks,” in Industrial Informatics, 2007 5th IEEE International Conference on, vol. 1, 2007, pp. 273–278.

[5] K. Koumpis, L. Hanna, M. Andersson, and M. Johansson, “Wireless industrial control and monitoring beyond cable replacement,” in Proc.2nd PROFIBUS Int. Conf., Coombe Abbey, Warwickshire, UK, 2005.

[6] S. Trikaliotis and A. Gnad, “Mapping wirelesshart into profinet and profibus fieldbusses,” in Emerging Technologies Factory Automation, 2009. ETFA 2009. IEEE Conference on, 2009, pp. 1–4.

[7] Siemens, “Profinet the industrial Ethernet standard for automation,” 8th IEEE International Workshop on Factory Communication Systems COMMUNICATION in AUTOMATION, 2007. [Online]. Available: http://wfcs2010.loria.fr/files/Siemens.pdf

[8] D. Miorandi, E. Uhlemann, S. Vitturi, and A. Willig, “Guest editorial: Special section on wireless technologies in factory and industrial automation, part i,” Industrial Informatics, IEEE Transactions on, vol. 3, no. 2, pp. 95–98, 2007.

[9] J. Antony, B. Mahato, S. Sharma, and G. Chitranshi, “A web plc using distributed web servers for data acquisition and control: Wed based plc,” in Information Science and Applications (ICISA), 2011 International Conference on, 2011, pp. 1–4.

[10] J.-S. Lee and P.-L. Hsu, “An improved evaluation of ladder logic diagrams and petri nets for the sequence controller design in manufacturing systems,” The International Journal of Advanced Manufacturing Technology, vol. 24, no. 3-4, pp. 279–287, 2004. [Online]. Available: http://dx.doi.org/10.1007/

[11] R. Zurawski and M. Zhou, “Petri nets and industrial applications: A tutorial,” Industrial Electronics, IEEE Transactions on, vol. 41, no. 6, pp.567–583, 1994.

[12] L. Gomes, A. Costa, J. Barros, R. Pais, T. Rodrigues, and R. Ferreira, “Petri net based building automation and monitoring system,” in Industrial Informatics, 2007 5th IEEE International Conference on, vol. 1, 2007, pp. 57–62.

[13] X. Fu, Z. Ma, Z. Yu, and G. Fu, “On Wireless sensor networks formal modeling based on Petri nets,” in Wireless Communications, Networking and Mobile Computing (WiCOM), 2011 7th International Conf.

[14] J. CERVANTES, “Representacion y aprendizaje de conocimiento con redes de petri difusas,” 2005.

[15] Corso, F.; Camargo, Y.; Ramirez, L., "Wireless Sensor System According to the Concept of IoT -Internet of Things-," Computational Science and Computational Intelligence (CSCI), 2014 International Conference on , vol.1, no., pp.52,58, 10-13 March 2014



www.elvedit.com 342

[16] Woped at http://www.woped.org/

[17] L. Palopoli, R. Passerone, and T. Rizano, “Scalable offline optimization of industrial Wireless sensor networks,” Industrial Informatics, IEEE Transactions on, vol. 7, no. 2, pp. 328–339, 2011.

[18] V. Jacobson, “Pathchar: A Tool to Infer Characteristics of Internet Paths,” ftp://ftp.ee.lbl.gov/pathchar/,

[19] Smokeping at http://oss.oetiker.ch/smokeping/

[20] N. V. Mnisi, O. Oyedapo, and A. Kurien, “Active throughput estimation using rtt of differing icmp packet sizes,” in Broadband Communications, Information Technology Biomedical Applications, 2008 Third International Conference on, 2008, pp. 480–485.

[21] Mnisi, N. V.; Oyedapo, O.J.; Kurien, A, "Active Throughput Estimation Using RTT of Differing ICMP Packet Sizes," Broadband Communications, Information Technology & Biomedical Applications, 2008 Third International Conference on , vol., no., pp.480,485, 23-26 Nov. 2008

[22] Y. Huang and G. Li, “Descriptive models for internet of things,” in Intelligent Control and Information Processing (ICICIP), 2010 International Conference on, 2010, pp. 483–486.

[23] Q. Zhu, R. Wang, Q. Chen, Y. Liu, and W. Qin, “Iot gateway: Bridgingwireless sensor networks into internet of things,” in Embedded and Ubiquitous Computing (EUC), 2010 IEEE/IFIP 8th International Conf.

[24] A. Williams. (2011, April) How the internet of things helps us understand radiation levels. [Online]. Available: http://readwrite.com/2011/04/01/ow-theinternet- of-things-help

[25] F. de la innovacion Bankinter, “El internet de las cosas.”

[26] C. Alcaraz, P. Najera, J. Lopez, and R. Roman, “Wireless sensor networks and the internet of things: Do we need a complete integration?” in 1st International workshop on the security of The internet of Things (SecIoT), 2010.

[27] A. Lakshmi, B. Sangeetha, A. Naveenkumar, B. Ganesh, and N. Bharathi, “Experimental validation of pid based cascade control system through scada-plc-opc interface,” in Computer Communication and Informatics (ICCCI), 2012 International Conference on, 2012, pp. 1–4.


Juan Felipe Corso Arias is M.Sc in Mechatronics from the Military Nueva Granada University and Electronic Engineer from the El Bosque University. He has been professor in the Networking field for the Military Nueva Granada University and in the Electronic field for the Catholic University of Colombia. Currently, he works as a SysAdmin for the Military Nueva Granada University. His research fields are related to networking, domotic and automation.



Yeison Julian Camargo Barajas is an Engineer in Telecommunications from the Military Nueva Granada University. He obtained his bachelor degree with a research work related to QoS multicast routing using artificial intelligence and a meritorious distinction for his research quality. He was granted with a scholarship by the National Department of science and technology to continue his research in the same university. He has worked for the industry in the networking and VoIP fields. His research fields are related to

Networking, Artificial Intelligence and Security.

Leonardo Juan Ramírez López is Chief of Division Technology Development and Innovation at the Nueva Granada Military University, Bogota - Colombia. Ph.D. Biomedical Engineering, University of Mogi das Cruzes, Sao Paulo, Brazil. Magister of Systems Engineering, National University of Colombia (2006). University Degree in Electronic Engineering (1997). Membership of International Society for Telemedicine and eHealth (ISfTeH) since 2010. IEEE member of EMBS Society since 2000. Associate Professor and leader of Telemedicine Group TIGUM-UMNG.


Analysis of Multiple String Pattern Matching Algorithms

Author

Akinul Islam Jony Department of Informatics, Technical University of Munich

[email protected] Munich, Germany

Abstract

Multiple string pattern matching is a basic problem in computer science and is used to locate all the appearances of a finite set of patterns inside an input text. It is widely used in many applications for searching, matching, filtering, and detecting a set of pattern. In this paper, to illustrate and for the better understanding of this particular problem, the widely used multiple string patterns matching algorithms have been analyzed and discussed. A theoretical and experimental result along with the analysis and discussion of the algorithms is presented as well in this paper. An extensive reference list is also included at the end of the paper.

Key Words

Algorithms, Multiple Pattern, String Matching, String Searching.

I. INTRODUCTION

String pattern matching or searching is the act of checking for the presence of the constituents of a given pattern in a given text where the pattern and the text are strings over some alphabet. It is an important component of many problems and it is used in many application such as text editing, data retrieval, data filtering (also called data mining) to find selected patterns, DNA sequence matching, detecting certain suspicious keywords in security applications, and of course, many other applications. The string searching or string matching problem consists of finding all occurrences of a set of pattern in a text, where the pattern and the text are strings over some alphabet.

Multiple string pattern matching problems has been a topic of intensive research that has resulted in several approaches for the solution such as multiple keyword generalization of Boyer-

Analysis of Multiple String Pattern Matching Algorithms Akinul Islam Jony


Moore algorithm [4], Boyer-Moore-Horspool algorithm [5] (which is simplified version of Boyer-Moore algorithm), Aho-Corasick algorithm [1], Commentz-Walter algorithm [2], Fan-Su algorithm [11] (which is a combination of Aho-Corasick and Boyer-Moore algorithms), Wu-Manber algorithm [3], and Set Backward Oracle Matching (SBOM) algorithm [12], [13] (which is the extension of the Backward Oracle Matching (BOM) algorithm [13], [14]). This paper mainly presents the analysis of mostly used algorithms for multiple string pattern matching problems: the Aho-Corasick algorithm, the Commentz-Walter algorithm, and the Wu-Manber algorithm. Experimental results of these algorithms are included for the analysis and discussion about multiple pattern matching problems. This paper also discusses the main theoretical results for each of the algorithm. The performance of each algorithm is shown against the length of pattern and the number of pattern in a pattern set. An extensive list of references is also presented at the end of this paper.

This paper structures as follows: Section II briefly describes the multiple pattern matching algorithms specifically Aho-Corasick, Commentz-Walter, and the Wu-Manber algorithm, Section III outline the experiment methodology, Section IV presents the experimental results on the multiple pattern matching algorithms, Section V presents the analysis and discussion on pattern matching problem based on the experimental results and existing works, and Section VI gives the conclusion of this paper.

II. MULTIPLE PATTERN MATCHING ALGORITHMS

This section presents the overview of most popular solutions for the multi-pattern matching problem: Aho-Corasick algorithm [1], Commentz-Walter algorithm [2], and Wu-Manber Algorithm [3].

A. Aho-Corasick algorithm

Aho-Corasick algorithm, a variant of the Knuth-Morris-Pratt algorithm [7], was the first algorithm to solve the multiple string pattern matching problems in linear time based on automata approach. This algorithm serves as the basis for the UNIX tool fgrep.

Aho-Corasick algorithm consists of two parts. In the first part they constructed a finite state pattern matching machine from the set of keywords and in the second part, the text string as input is applied to the pattern matching machine. The machine signals whenever it has found a match for a keyword (pattern). The pattern matching machine consists of a set of states and each state is represented by a number. The behavior of the pattern matching machine is dictated by three functions: a goto function g, a failure function f, and an output function output. Figure 1 shows a sample pattern matching machine for the set of patterns, p = {he, she, his, hers} [1].

The goto function g maps a pair consisting of a state and an input symbol into a state or the message fail. The failure function f maps a state into a state. The failure function is consulted whenever the goto function reports fail. The output function of certain states indicates that a set of keywords has been found.



www.elvedit.com 346

FIGURE 1: A SAMPLE PATTERN MATCHING MACHINE.

The construction of Aho-Corasick automaton machine takes running time linear in the sum of the lengths of all patterns/ keywords. This involves building the pattern tree (keyword tree) for the set of pattern and then converting the tree to an automaton (pattern matching machine) by defining the functions g (goto function), f (failure function), and output function for labeling states with the keyword(s) matched. The space or memory requirements of the Aho-Corasick algorithm can be quite large depending on the pattern set and also the length of each pattern in a pattern set. The matching process is simply stepping through the input characters one at a time and checks if there is any matching. Each step in pattern matching machine happens in constant time. So, the Aho-Corasick matcher always operates in O(n) running time.

B. Commentz-Walter algorithm

Commentz-Walter presented an algorithm for the multi-pattern matching problem that combines the Boyer-Moore technique with the Aho-Corasick algorithm. Commentz-Walter combines the filtering functions of the single pattern matching Boyer-Moore algorithm and a suffix automaton to search for the occurrence of multiple patterns in an input string. The tree used by Commentz-Walter is similar to that of Aho-Corasick’ pattern matching machine but is created from the reversed patterns. A sample reversed pattern tree is shown in figure 2.

FIGURE 2: A SAMPLE REVERSED PATTERN TREE



The original paper presented two versions of the algorithm. In all version of the algorithm a common program skeleton is used with different shift function. The Commentz-Walter algorithm is also consists of two phase: pre-computing phase and matching phase. The pre-computation phase of algorithm is responsible for creating a pattern tree by using the reversed pattern (see figure 2). The matching phase of the Commentz-Walter algorithm is combination of two ideas. One is from the ideas of Aho-Corasick’ finite automata technique (in pattern tree) and another one is from the Boyer-Moore shifting techniques (in right-to-left matching). In this algorithm a match is conducted by scanning backwards through the input string. At the point of mismatch some number of characters about the input string is known (that is, the number of characters that ware matched before the mismatch) and this information then is used as an index. The index is used in a pre-computed table to determine a distance which is later helps to shifting before commencing the next match attempt.

C. Wu-Manber Algorithm

Wu-Manber algorithm is a simple variant of the Boyer-Moore algorithm that uses the bad-character shift for multiple pattern matching. They actually come to the idea of this algorithm after making a UNIX based tool agrep [10] which was for searching many patterns in files. To improve the performance, they come through a unique idea, that is, their algorithms looks at block of text instead of single character. So, they consider both pattern and text as blocks of size B instead of single characters. As recommended in their paper [3], in practice B could be equal to 2 for a small pattern set size or to 3 otherwise.

The operational process of the WM algorithm includes two phases. In first phase which is called preprocessing phase, three tables are constructed by the patterns, the SHIFT, the PREFIX and the HASH tables. The SHIFT table stores the shift values of the block characters that determine the safe shifting of characters during the searching phase. If a block of B characters does not occur in any pattern, then the shift value for that block assigns to the maximum value, which is m − B + 1. The HASH table stores hashed values (h) of B characters suffix of each pattern while the PREFIX stores hashed values (h′) of B’ characters prefix of a list of patterns that they have the same suffix.

The second part of the algorithm is the searching phase. During this phase, the algorithm is searching for the occurrences of all patterns in the input text with the assistant of the three tables that have been created by the previous state. Firstly, a hash value (h) for the block of B characters is calculated into the current search window and the shift value for that is checked (SHIFT[h]). If the shift value is greater than zero, then the current search window is shifted by SHIFT[h] positions, or else there is a potential matching and the tables HASH and PREFIX should be considered in order to validate the matching.

The first thing is to compute the minimum length of a pattern, call it m, and consider only the first m characters of each pattern. This requirement is crucial to the efficiency of the algorithm. If one of the patterns is very short, say of length 2, then it is not possible to shift by more than 2, so having short patterns inherently makes this approach less efficient. The expected running time



www.elvedit.com 348

complexity of the main matching phase/ searching phase is shown by Wu-Manber to be less than linear in n (the length of the input text) [3].

III. EXPERIMENT METHODOLOGY

To evaluate the performance of the multiple pattern matching algorithms, the practical running time is used as a measure. Practical running time is the total time an algorithm needs to find all occurrences of a pattern set in an input text including any preprocessing time. In the experiment, English text is considered as an input text where the pattern set is chosen randomly for the searching/ matching process. The input text used in this experiment is consists of 477,048 characters excluding spaces and it contains 99,449 words in total. For the implementation of these algorithms JAVA is used as a programming language. All the experiments are run on a computer which has a 2.20 GHz Intel Core 2 Due processor, 4 GB RAM, and 64-bit Windows 7 Operating System.

IV. EXPERIMENTAL RESULTS

In this section of the paper the experimental results of the algorithms are presented. The performance of the algorithms is shown by measuring the running time against the number of pattern and length of pattern (pattern size) in a pattern set. The table 1 and figure 3 shows the running time of Aho-Corasick algorithm with different number of patterns (10 to 500 patterns) but the minimum length of pattern is 2. In Aho-Corasick algorithm, if the number of pattern is increases, the running time is also increase.

Table 2 and figure 4 also shows the running time of Aho-Corasick algorithm but in this time with different length of pattern (length 3 to 17) and fixed number of pattern (in this case 10). In Aho-Corasick algorithm, if the length of pattern is increases, the running time is also increase.

TABLE 1: RUNNING TIME OF AHO-CORASICK ALGORITHM DEPENDING ON

NUMBER OF PATTERNS Number

of pattern

Minimum length of pattern

Running time (ms)

10 2 32 50 2 39 100 2 48 150 2 48 200 2 51 250 2 60 300 2 63 350 2 63 400 2 64 450 2 69 500 2 73

FIGURE 3: RUNNING TIME OF AHO-CORASICK ALGORITHM DEPENDING ON NUMBER OF PATTERNS

30354045505560657075

10 50 100 150 200 250 300 350 400 450 500

Runn

ing T

ime

(ms)

Number of Pattern (10 to 500)



But it has better performance with the larger length of pattern. As we can see that running time in Aho-Corasick algorithm does not change too much with the larger length of pattern.

The Table 3 and Figure 5 shows the running time of Commentz-Walter algorithms with different number of patterns (10 to 500 patterns) but the minimum length of pattern is 2. Like in the Aho-Corasick algorithm, the running time of Commentz-Walter algorithm is also increasing with the number of pattern increases.

TABLE 2: RUNNING TIME OF AHO-CORASICK ALGORITHM DEPENDING

ON LENGTH OF PATTERNS Number of pattern

Length of pattern

Running time (ms)

10 3 32 10 4 33 10 5 33 10 6 35 10 7 39 10 8 39 10 9 39 10 10 47 10 11 47 10 12 47 10 13 48 10 14 48 10 15 48 10 16 48 10 17 49

FIGURE 4: RUNNING TIME OF AHO-CORASICK ALGORITHM DEPENDING ON LENGTH OF PATTERNS

30

35

40

45

50

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Runn

ing T

ime

(ms)

Length of Pattern (3-17)

TABLE 3: RUNNING TIME OF COMMENTZ-WALTER ALGORITHM DEPENDING ON NUMBER

OF PATTERNS Number of

pattern Minimum length

of pattern Running time (ms)

10 2 30 50 2 38 100 2 46 150 2 48 200 2 49 250 2 54 300 2 58 350 2 60 400 2 62 450 2 67 500 2 70

FIGURE 5: RUNNING TIME OF COMMENTZ-WALTER ALGORITHM DEPENDING ON NUMBER OF PATTERNS

30354045505560657075

10 50 100

150

200

250

300

350

400

450

500

Runn

ing T

ime

(ms)




www.elvedit.com 350

Table 4 and figure 6 shows the running time of Commentz-Walter algorithm with different length of pattern (length 3 to 17) but fixed number of pattern (in this case 10). In Commentz-Walter algorithm, if the length of pattern is increases, the running time is also increase. But the running time of Commentz-Walter algorithms improved approximately linearly with increasing length of the shortest pattern in the pattern set.

TABLE 4: RUNNING TIME OF COMMENTZ-WALTER ALGORITHM DEPENDING ON

LENGTH OF PATTERNS Number

of pattern Length of pattern

Running time (ms)

10 3 31 10 4 33 10 5 33 10 6 34 10 7 36 10 8 37 10 9 39 10 10 41 10 11 41 10 12 43 10 13 44 10 14 44 10 15 46 10 16 47 10 17 47

FIGURE 6: RUNNING TIME OF COMMENTZ-WALTER ALGORITHM DEPENDING ON LENGTH OF PATTERNS

30

35

40

45

50

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Runn

ing T

ime

(ms)

Length of Pattern (3-17)

TABLE 5: RUNNING TIME OF WU-MANBER ALGORITHM DEPENDING ON NUMBER OF

PATTERNS Number of

pattern Minimum length

of pattern Running time (ms)

10 3 15 50 3 18 100 3 19 150 3 21 200 3 21 250 3 24 300 3 25 350 3 27 400 3 29 450 3 33 500 3 34

FIGURE 7: RUNNING TIME OF WU-MANBER ALGORITHM DEPENDING ON NUMBER OF PATTERNS

10

15

20

25

30

35

40

10 50 100

150

200

250

300

350

400

450

500

Runn

ing T

ime

(ms)




The table 5 and figure 7 shows the running time of Wu-Manber algorithm with different number of patterns (10 to 500 patterns) but the minimum length of pattern is 3. In Wu-Manber algorithm, if the number of pattern is increases, the running time is also increase. But the performance of this algorithm is better than the Aho-Corasick algorithm because, Wu-Manber algorithm use block of character shifting while searching for a set of pattern in a given text. Furthermore, as Aho-Corasick and Commentz-Walter algorithms are based on automata approach, and hence, they consume more memory than Wu-Manber algorithm.

V. ANALYSIS & DISCUSSION

A linear time algorithm for multiple patterns matching problem proposed by Aho and Corasick [1] is optimal in worst case but Boyer and Moore [4] demonstrated an algorithm where they showed that it is possible to skip a large portion of the text while searching for certain patterns. Eventually, this is working faster than linear algorithm in the average case. The Commentz-Walter algorithm [2] combines the idea of Boyer and Moore technique with Aho-Corasick algorithm for multiple patterns matching problem which is substantially faster than the Aho-Corasick algorithm in practice. It uses the idea of Boyer Moore technique to skip a large portion of the text while searching and as a result leading to faster than linear time algorithms in the average case. There has another algorithm proposed by Baeza-Yates [6] which also combines the idea of Boyer-Moore-Horspool algorithm [5] (which is a slight variation of the classical Boyer-Moore algorithm) with the Aho-Corasick algorithm. Whereas, Wu-Manber algorithm is the most efficient algorithm under some scenarios such as, long random patterns, low matching rate, and low memory requirement. However, Aho-Corasick performance does not suffer great decline when comparing with others as it is a linear time searching algorithm in worst case.

Independent from the pattern set size, searching time complexity for Aho-Corasick algorithm is O(n) but when pattern set size increase, the memory consumption increased drastically and also the time consumption increased. The performance of Commentz-Walter algorithms declined with increasing number of pattern in a pattern set (pattern set size). But the performance of Commentz-Walter algorithms improved approximately linearly with increasing length of the shortest keyword/ pattern in the pattern set. Moreover, in [8, pp. 281], [9], A.V. Aho states that,

“In practice, with small numbers of keywords, the Boyer-Moore aspects of the Commentz-Walter algorithm can make it faster than the Aho-Corasick algorithm, but with larger numbers of keywords the Aho-Corasick algorithm has a slight edge.”

This paper also found the above statement correct with the presented experimental result and analysis. But the Aho-Corasick and Commentz-Walter algorithms consume lots of memory because in the preprocessing stage both these algorithms use the automata data structure whereas Wu-Manber algorithm consume much less memory than these two algorithms.



www.elvedit.com 352

VI. CONCLUSION

The algorithms that offer the solution for the multi-pattern matching problem, among them Aho-Corasick, Commentz-Walter, and Wu-Manber algorithms are very popular solutions. A comprehensive analysis and discussion of these selected algorithms as a state-of-the-art along with some experimental results is covered in this article. This paper has presented the analysis and discussion of the algorithms in order to understand the multiple pattern matching problem in an easier manner.

The Aho-Corasick algorithm considers as a classic solution and core element of many other pattern matching algorithms. Also it has been used in many other applications. As it is a linear time algorithm, it is considered very useful solution for multiple pattern matching problems. On the other side, Commentz-Walter algorithm seems to be the first sub-linear running time algorithm for multiple-pattern matching problems in average case by using a sifting technique where a large portion of the text is skipped while searching. The Wu-Manber algorithm has excellent average case performance because of the successful use of shifting operation as a block of characters. However, Wu-Manber algorithm has minimum length problem. If the minimum length of pattern is less, then it is not as efficient as it should be. So it would be an optimization area for Wu-Manber algorithm. For the Aho-Corasick and Commentz-Walter algorithms, memory compression would be a good optimizing area as they consume lots of memory. Also subset division of pattern set could be another way for the optimization of Aho-Corasick algorithm.

This paper mainly covers the analysis and discussion among Aho-Corasick, Commentz-Walter, and Wu-Manber algorithms for multiple string pattern matching problems. A comprehensive study on all the existing algorithms of multiple pattern matching problems would be a very demanding material in the research area of multiple pattern matching problems.

REFERENCES

[1] Aho, Alfred V. & Corasick, Margaret J. (1975). Efficient string matching: an aid to bibliographic search. Communications of the ACM, 18, 333-340.

[2] Commentz-Walter, Beate. (1979). A string matching algorithm fast on the average. Automata Languages and Programming, 6, 118-132.

[3] Wu, Sun & Manber, Udi. (1994). A fast algorithm for multi-pattern searching. Technical Report TR-94-17, University of Arizona.

[4] Boyer, R. S. & Moore, J. S. (1977). A fast string searching algorithm. Communications of the ACM, 20, 762-772.

[5] Horspool, N. (1980). Practical fast searching in strings. Software: Practice and Experience, 10, 501-506.

[6] Baeza-Yates, R. A. (1989). ‘Improved string searching, Software - Practice and Experience, 19, 257-271.

[7] Knuth, Donald E., Morris, James H., Pratt, Vaughan R. (1974). Fast pattern matching in strings. Technical Report CS440, Computer Science Department, Stanford University, Stanford, California.



[8] Aho, A. V. (1990). Algorithms for finding patterns in strings. Handbook of Theoretical Computer Science In J. van Leeuwen (Ed.), (pp. 257-300). North-Holland, Amsterdam.

[9] Waston, B. W. (1994). The performance of single-keyword and multiple keyword pattern matching algorithms.

[10] Wu, S. & Manber, U. (1992). Agrep – a fast approximate pattern-matching tool. In Proceedings USENIX Winter 1992 Technical Conference, (pp. 153–162), San Francisco, CA.

[11] Fan, J.-J. & Su, K.-Y. (1993). An efficient algorithm for matching multiple patterns. IEEE

Transactions on Knowledge and Data Engineering, 5(2), 339–351.

[12] Navarro, G. & Raffinot, M. (2002). Flexible Pattern Matching in Strings, Practical Online Search Algorithms for Texts and Biological Sequences. Cambridge University Press, Cambridge, UK.

[13] Allauzen, C. & Raffinot, M. (1999). Oracle des facteurs d’un ensemble de mots. Technical Report IGM 99-11, Institut Gaspard Monge, Universit´e de Marne-la-Vall´ee,France.

[14] Allauzen, C., Crochemore, M., Raffinot, M. (1999). Factor oracle: A new structure for pattern matching. In J. Pavelka, G. Tel, and M. Bartosek, (Ed.), Theory and Practice of Informatics (Brno, 1999), volume 1725 of Lecture Notes in Computer Science, pp. 291–306. Springer-Verlag. In Proceedings of the 26th Seminar on Current Trends in Theory and Practice of Informatics, Milovy, Czech Republic.


Akinul Islam Jony born in Dhaka, Bangladesh. He received his M.Sc. degree in Informatics at Technical University of Munich (TUM) in Germany. Previously he completed his B.Sc. degree in Computer Science at American International University – Bangladesh (AIUB) and Master degree in Information Technology at University of Dhaka (DU) in Bangladesh. His current research interest includes algorithms, service-oriented computing, distributed middleware system, and ubiquitous computing.


Nonlinearity Compensation for High Power Amplifiers Based on Look-Up Table Method

for OFDM Transmitters

Authors

Maryam Sajedin Department of Electrical Engineering, Fars Science and Research Branch, Islamic Azad University,

[email protected] Fars, Iran

Ayaz Ghorbani Department of Electrical Engineering, Amirkabir University of Technology

[email protected] Tehran, P.O. Box: 15875-4413,

Iran

Hamid Reza Amin Davar Department of Electrical Engineering, Amirkabir University of Technology

[email protected] Tehran, P.O. Box: 15875-4413, Iran

Abstract

The OFDM is generally known as an effective technique for high Bit-rate applications. In OFDM systems, the combination of different signals with different phase and amplitude give a large dynamic range that is used to be characterized by a high PAPR. To obtain maximum efficiency Power Amplifier should be driven near the saturation region, but since the OFDM signal has high PAPR, this power amplifier will cross over to the nonlinear region causing serious in band distortion, and operation in nonlinear mode reduces performances of the output signal. To compensate for this distortion, liberalizers are proposes to utilize a digital pre-distortion of baseband signals, which is efficient and illustrates a high performance for linearization of OFDM transmitters. This paper presents an adaptive digital pre-distortion techniques based on Look Up Table (LUT) method which will result in cancellation of nonlinear distortion appearing in power amplifier through Advanced Design System(ADS) software. It is shown that the new simplified structure exhibits fast convergence and LUT pre-distorter can effectively suppress the spectrum regrowth caused by the dynamic nonlinearity of power amplifier.

Key Words

About four key words or phrases in alphabetical order, separated by commas.

Nonlinearity Compensation for High Power Amplifiers Based on Look-Up Table Method for OFDM Transmitters Maryam Maryam Sajedin, Ayaz Ghorbani and Hamid Reza Amin Davar


I. INTRODUCTION In recent years OFDM has attracted a great deal of attention for digital terrestrial

broadcasting and mainly in 4G technology considered for modulation. The OFDM is a combination of modulation and multiplexing and it is a multicarrier transmission technique too. It uses the spectrum so efficiently by spacing the channels much closer together [1]. Also, it can reduce the frequency selectivity of the channel taking advantage a simple one-tap equalizer [2]. The OFDM signal is robust against multipath fading and impulsive noise [3] and behaves like a Gaussian random process. Furthermore, the inter-symbol interference (ISI) in OFDM can be easily prevented by inserting a guard interval before each transmitted block, longer than the largest delay spread [4]. The OFDM structure requires a summation of a large number of subcarrier for multicarrier modulation and as a result of this summation large signal envelope fluctuations occur.

However, one of the important problems associated with OFDM is its high peak to average power ratio [5] which requires large power back off for linear operation of PA, resulting in a low average efficiency [6]. This nonlinear distortion causes serious in band distortion as well as adjacent channel interference due to spectrum regrowth in the transmitted signal. High Power Amplifier (HPA) working and performance plays the great role in OFDM wireless system [7]. Real PA has a nonlinear response that creates in-band and out-of-band distortion that not only reduces the system performance but also creates interference on adjacent channels (ACI). The nonlinear effects on the transmitted OFDM signal are: spectral-spreading of the subcarriers warping of the signal constellation in each sub channel. Nonlinear amplifiers are characterized by measurement of their AM/AM (amplitude dependent gain) and AM/PM (amplitude dependent phase shift) function in either polar or quadrate form [8]. To obtain maximum efficiency the power amplifier should be driven near the saturation region, but since the OFDM signal has high PAPR these power amplifier will cross over to the nonlinear region causing serious in band distortion. Therefore linearizing techniques should be introduced to minimize the output distortion. The most rapidly developing linearization technique is digital pre-distortion (DPD), this is a popular and reliable technique that allows minimizing output distortion and spectral regrowth [9]. The most developed DPD methods are Look up Table and polynomial [10].

In this paper we the authors characterize the performance of the OFDM in the presence of a High Power Amplifier .Taking advantage of the Fourier transformation the output correlation function can provide information on the output power spectral density (PSD) .An adaptive digital baseband compensator based on the LUT (Look-Up Table ) method is proposed to overcome the nonlinear distortion .We demonstrate that our simplified scheme exhibits fast convergence. Section 2 briefly describes the concepts of OFDM transceiver, introduces the power amplifier model and the effects of HPA nonlinear distortion on OFDM signal and LUT method. The computer simulations and the experimental results are given in section 3 and finally section 4 concludes the paper.

II. MATERIAL AND METHODS

A. Signal Model



www.elvedit.com 356

Figure 1 illustrates a baseband transceiver structure for OFDM using the Fourier transform for modulation and demodulation. Here the serial data stream is mapped to complex data symbols (QAM) with a symbol rate of . The data is then demultiplexed by a serial to parallel converter resulting in a block of N complex symbols, x tox . The parallel samples are then passed through an N point IFFT (in this case no oversampling is assumed) with a rectangular window of length N. T . Resulting in complex samples x to x assuming the incoming complex data is random it follows that the IFFT is a set of N independent random complex sinusoids summed together. The samples, x to x are then converted back into a serial data stream producing a baseband OFDM transmit symbol of length t = N. T .

FIGURE 1: THE BASIC OFDM TRANSMITTER AND RECEIVER PAIR UTILIZING FOURIER TRANSFORM A Cyclic Prefix (CP), which is a copy of the last part of samples is appended to the front of

serial data stream before Radio frequency up conversion and transmission. It combats the disrupting effects of the channel which introduce Inter System Interference. In the receiver the whole process is reversed to recover the transmitter data, the CP is removed prior to the FFT which reverses the effect of the IFFT. The complex symbols at the out of the FFT , y ……y are then decoded and the original bit stream recovered [11].

B. HPA Models

The following equations describe the algorithm used for PA model. The input signal V (t) is represented by its inphase and quadrature components about its carrier frequency.



푉 (푡) = 푅푒{푉 (푡)푒 } ,푉 (푡) = 푉 (푡) + 푗푉 (푡) (1)

The output signal V (t) is then given by the equation

푉 (푡) = 푅푒{푎푔 푉 (푡)푒 } (2)

Where a denotes the gain of the component . {If the input is a baseband timed signal, then only the real part of the Gain is used for a}. g denotes the gain compression factor as determined by the gain compression parameters, ( GCType(Gaincompressiontype) , TOIout(Thirdorderinterceptpower) , dBc1out(1dBgaincompressionpower) , PSat(Saturationpower), GCSat(Gaincompressionatsaturation)) . All gain compression characteristics, are modeled using a polynomial expression up to the saturation point; after this point, output power is held constant for increasing input power. The gain compression expression for nonlinear models is defined with a nonlinear amplitude characteristic. When GCType = PSat + GCSat + TOI, then the g comp factor is due only to the output third-order intercept point TOIout, output saturated power PSat, and the gain compression at saturation GCSat, where (3≤GCSat≤7, and (TOI−10+0.5(GCSat-1)) ≤ PSat≤(TOI−4).

FIGURE 2: ILLUSTRATION OF THE THIRD ORDER INTERCEPTS POINTS

In figure (2), it can be shown that the slope of the linear gain for input and output powers in dBs is unity, likewise the slope of the third gain of the third order IMD component is 3, the point where the third order line intersects with the linear gain line is the third order intercept point. Nonlinear models TOI through PSat + GCSat + TOI + dBc1mathematical gain model,



www.elvedit.com 358

푉 (푉 ) = 푎 푉 + 푎 푉 + 푎 푣 + ⋯ (3)

Where V presents input signal voltage, V is output signal voltage, a illustrates small signal gain, a shows third-order gain coefficient and a , presents higher odd-order gain coefficients. The gain compression expression for nonlinear model, in general, has both amplitude and phase changes versus increasing input power. Ref.[12] have shown that only the odd components of the nonlinear model bring distortion to the fundamental signal. When a single carrier input signal, is substituted into above formula (1), the output waveform will contain the original sine wave and harmonic distortion products [12] the harmonics can be eliminated by filtering and do not pose a problem except for wideband communication application requiring wide bandwidth. However, when more than one carrier is present, additional new signals known intermodulation distortion (IMD) are produced in the vicinity of input signals. Filtering cannot easily eliminate IMD products, as these are located on the same frequency or near to the desired input signal [13].

(a) (b)

FIGURE 3: (A) AM/AM AND (B) AM/PM TRANSFER CURVES

Figure (3) presents the distortion characteristics of power, use of a exact approximation of these nonlinearities allows a linearization good enough. The accuracy and efficiency of the pre-distortion rely strongly on modeling of the true nonlinearities.

C. The Effects of HPA Nonlinear Distortion on OFDM Signals

HPA nonlinearity may have bad influence on OFDM signal mainly on two aspects: a out of band distortion, which will cause the OFDM power spectrum distortion i.e. the spectral spreading of the amplified signal and introduce ACI, as is shown in figure 4 right, requirements on ACI for RF systems are very strict especially with large number of subscribers, therefore, it is of great importance to distortion, which may disturb the OFDM constellations and result in BER performance degradation. Figure 4 left; show that the spectrums of OFDM signals through HPA



have spectrum re-growth distortion.

Regulatory bodies specify power spectral density masks which define the maximum allowable adjacent channel interference (ACI) levels. In order to meet the regulatory mask, at least a 20 dB improvement in the intermodulation products is required.to satisfy these requirements, linearizing techniques should be introduced to minimize the output distortion. the Digital Pre-distortion (DPD) is one of the promising linearization techniques ,since it allows the use of well

developed digital signal processing techniques in the Baseband [14, 15].

FIGURE 4: LEFT: THE POWER SPECTRUM DENSITY (PSD) REFERENCE OFDM SIGNAL SPECTRUM , RIGHT:

DISTORTED OFDM SIGNAL SPECTRUM

D. The Adaptive Digital Predistortion System

Pre-distortion techniques have proposed as a potential solution to overcome the nonlinear effects [15], it is equivalent to a nonlinear circuit with gain expansion response that is the inverse of the PA gain compression (AM/AM) response and a phase rotation that is the negative of the PA phase rotation (AM/PM), when combining the pre-distortion with the PA that can compensate the distortion generated by the nonlinear amplifier. Pre-distortion is widely used as a method in which signal processing is applied to the time signal before it is input to the amplifier [16]. A predistorter can successfully correct distortion up to the full saturation level of the amplifier. Alternative LUT adaption techniques with low complexity and low memory requirement proposed in the literature. Building a LUT predistorter from a set of stored input and output complex envelope samples is a trivial process. A block diagram of an adaptive digital predistortion system is shown in figure 5. The predistorter consists of a complex gain adjuster which controls the amplitude and phase of the input signal. The amount of predistortion is



www.elvedit.com 360

controlled by the entries of a Look-up Table that interpolate the AM/AM and AM/PM nonlinearities of the power amplifier. The feedback path samples the distorted signal for which the DSP adjusts the Look-up Table entries so as to minimize the level of distortion.

FIGURE 5: DIGITAL PREDISTORTION BLOCK DIAGRAM

Notably, the nonlinear distortion is determines by the signal envelope [17]. Thus, using the input signal envelope would be much more efficient way to addressing of the predistortion Look Up Table. The LUT coefficients implement the predisotion function. The adaptation algorithm determines the values of the coefficients by comparing the feedback signal and a delay version of the input signal. The size of the LUT employed determines the number of points at which the predistortion function is calculated [18]. The LUT is in fact implemented by two RAMs of which the first determines the magnitude of the complex gain, whereas the second one determines the phase shift. It has been shown that if the input signal is Gaussian the best Look Up Table address spacing is linear [19]. It’s mentioned that the OFDM signal being the sum of a large number of QAM or QPS modulated carriers is approximately Gaussian.



FIGURE 6: THE ALGORITHM OF LUT

Figure 6 describes the basic algorithm implemented in the LUT design. The incoming complex samples in I and Q, have correction factors applied from the LUT and sent to module. The addresses for the LUT are derived from the input power. The LUT must contain two values for each location the real part and the imaginary part Q. in the module ,samples are unconverted and sent to the PA . In the feedback loop, the output of the PA is downconverted—transformed to polar form—and compared with the delayed version of the input to the predistorter in polar form. This error is then used to update the values currently stored in the LUT. The LUT address is derived from the input power. Hence this algorithm is only able to correct for phase and magnitude error that are purely a function of the current input power. The LUT coefficient is fed into the predistorter, which reads an appropriate correction value (LUT value) from the LUT and uses it to modify the input data .the resultant modified coefficient is referred to as predistorted data.

The basic idea of determine the LUT coefficients fairly straightforward. By considering the scenario depicted in Figure 7, let r be the amplitude of the input signal. The desired output is known from the linear response. This value is used to search through the output characteristic of the amplifier. The value r is the desired output amplitude; from which the proper input amplitude to the amplifier is determinedr . The original input r ,amplitude is adjusted to produce r [20]. Thenr should produce the correct output amplitude to give the overall predistorter-power amplifier chain a linear response. Although not shownr , is also used to determine Φ to predistort the phase. Phase is not always corrected, but if it is, it can also compensate for any quadrature modulation errors in addition to the amplifier errors [21]. Note that if the desired output amplitude is beyond the saturation limit of the amplifier, the corresponding r will not be able to fully correct for the nonlinearity.



www.elvedit.com 362

FIGURE7. THE CONCEPTUAL MODEL OF PREDISTORTION.

Values of rversus A(r) and r versus ∆Φ(r) are stored in Look Up Tables. For every complex input r, the pre-distorter Look up the desired output level in A(r) and applies a correction factor based upon A(r ) to produce r , next r is compared to the closet value of r in the phase table and the predistorter applies the correction −∆Φ(r). The resulting output (and thus the input to the amplifier), is a pre-distorted sample [22]. The delay in the feedback path is estimated by calculating the correlation between the magnitude of the input signal and the magnitude of the feedback signal. The benefit of using the signal magnitude is that it does not require phase synchronization in the feedback path.

III. THE EXPERIMENTS AND RESULT

In this section the authors simulate a real digital predistorter based on the complex gain lookup table technique. In order to demonstrate linearizing performance of the baseband predistorter ,Advanced Design System (ADS) simulator were carried out with a OFDM signal. We also used the Linearization Design Guide, which provides a complete tool kit to interactively explore dynamic linearization systems at the top level as part of an integrated design process. Adaptation using the digital predistorter is very rapid. Figure 8 shows the schematic of the digital predistorter wherein all subsystems have been implemented in the ADS Agilent Ptolemy simulation. The main ADS digital circuit uses a data flow simulation controller in order to execute simulations.



FIGURE 8: PROPOSED LUT-BASED ADAPTIVE DIGITAL PREDISTORTER

This platform has been implemented using Agilent ADS software. Because of each block has been implemented at component level, so details of each component have not being given deliberately. The LUT address generation component translates the magnitude of the baseband input signal into a LUT address using power addressing schemes [23]. The LUT is implemented using the LUT_RAM and the number of entries in the LUT is taken as 256.

The simulations have been performed in the baseband domain. In figure 9, the simulated output spectra of the linearly amplified input signal, with the PA output, with and without predistortion are reported. Without predistortion a visible spectral re-growth is present, while using the proposed baseband predistortion the input and output spectra appropriately scale by a constant factor. The PA has been driven near the saturation. The spectral densities have been normalized with the maximum power of the desired output signal; we can appreciate the significance of the cancellation of PA memory in reducing in-band distortion.

Finally, we can determine the performance of our digital predistortion circuit. Figure 10 shows the output from the digital baseband predistorter once the LUT entries have adapted. We can observe the spectral growth that occurs using a predistorter. The adjacent channel power is spread over a wider bandwidth but the mask requirements can be meet. Approximately 20 dB of distortion correction is achieved.



www.elvedit.com 364

FIGURE 9: PA OUTPUT WITH AND WITHOUT DP OF THE MODELED PA IN ADS (RED: WITHOUT PREDISTORTION,

BLUE: WITH PREDISTORTION).

FIGURE 10: AM/AM CHARACTERIZATION; AND AM/PM CHARACTERIZATION

-10 -5 0 5 10-15 15

-120

-100

-80

-60

-40

-20

-140

0

Frequency (MHz)

Po

wer

(dB

m)

PA Output

Final Coefficients (iteration 6)

0.90

0.95

1.00

1.05

1.10

1.15

0.85

1.20

Mag

nitu

de

MagMarker

MagMarkerindep(MagMarker)=mag(LUTFinal)=1.813995

255

32 64 96 128 160 192 2240 256

-0.02

-0.01

0.00

0.01

0.02

-0.03

0.03

LUT Entry

Ph

ase

(deg

rees

)



The AM/AM and AM/PM transfer characteristic simulation curves derived from the polynomial HPA model are present in figure 10, which confirms the removal of the nonlinear distortion with memory effects by the LUT PD.

IV. CONCLUSION

In this paper, the effects of nonlinearities in the power amplifier over OFDM systems were investigated, it is noticed that the effects of nonlinearity of the high power amplifier depends upon the type of modulation used in OFDM system. This study has proposed a baseband digital compensation method for nonlinear distortions in digital transmitters. Performances of the digital predistortion circuit have been investigated. The implemented predistorter uses two LUTs containing the real and imaginary part of the adaptive predistortion function. LUT size, indexing, interpolation and update are important factors in the design of a digital predistorter. the LUT configuration reduces complexity in the implementation and permits in order to meet the tradeoff between complexity scalability and PD accuracy. The proposed design shows better performance in terms of improving ACPR, and easy to implement.

REFERENCES [1] F. H. Gregorio and T. I. Laakso.(2005). The performance of OFDM-SDMA systems with power

amplifier nonlinearities. Proceedings of the 2005 finnish signal symposium , Kuopio ,Finland.

[2] M. Alard and R. Lassale .(1987) . Principles of modulation and channel coding for digital broadcasting for mobile recivers . EBU Tech.Rev.,no224 , 3-25

[3] L. J. Cimini . (1985). Analysis and simulation of a digital mobile channel using orthogonal frequency division multiplexing. IEEE Trans, Commun , Com-33 , 665-667

[4] Vivek Ashok Bohara, See Ho Ting. (2008). Analysis of OFDM signals in nonlinear high power amplifier with memory .IEEE Communication in the ICC 2008 preceeding , 3653-3657

[5] Dytro Bonder, Djuradj Budmir and Boris Shelkovnikor. (2008) . Linearization of power amplifier by baseband digital predistortion for OFDM transmitter. INT.Crimean Conference Microwave & Telecommunication Technology , 270-272

[6] Jinho Jeong .(2012). New digital predistortion technique of RF power amplifier for wideband OFDM signal .IEICE Electronics Express ,Vol 9.No.5, 326-332

[7] Tushar Kanti and Monir Morshed.(2013). High power amplifier effects analysis for OFDM system. International Journal of Science , Engineering and Technology Research ,(IJSETR) ,Vol.2 , Issue 5 , 1119-1121

[8] Amanjot Singh ,Hardeep Kaur. (2012). nonlinearity analysis of power amplifier in OFDM system. International Journal of Computer Applications ,Vol .37 , 37-41



www.elvedit.com 366

[9] Bo Ai , Member, IEEE , Zhi –Xing Yang , Chang – Yong Pan, Shi –Gang Tang and Tao –Tao Zhang . (2007). Analysis on LUT based predistortion method for HPA with memory .IEEE Transaction on Broadcasting , Vol.53 , No.1 , 127-129

[10] Dmytro Bondar, Djuradj Budimir and Boris shelkonvniko. (2008). Linearization of power amplifiers by basedband digital predistortion for OFDM transmitters. IEEE Microwave & Telecommunication Technology,270-271

[11] Gavin Hill. (2011). Peak power reduction in orthogonal frequency division multiplexing transmitters. Victoria University of Technology ,Thesis submitted in fulfilment of the requirement for the degree of doctor of philosophy .

[12] Stevens Creek Blvd., Santa Clara. (2011). Advanced Design System 2011.01 - Timed Components. Agilent Technologies.

[13] Sangeeta Bawa ,Maninder Pal ,Jyoti Gupta. (2013). Predistortion based linearization technique for power amplifiers of wideband communication systems. International Journal of Science & Engineering Reserch ,Vol4 ,Issue 5

[14] ]AIBO , YAG , ZHI –XING, PAN CHANG –YONG, ZHANG TAO –TAO , WANG YONG and GE JIAN HUA. (2006). Improve LUT Technique for HPA Nonlinear predistortion in OFDM System. Wireless personal Communication , No 38 , 495-507

[15] Amanjot Singh ,Hardeep Kaur. (2012). Non Linearity Analysis of High Power Amplifier in OFDM system. International Journal of Computer Aplication ,Vol 37, NO.2

[16] Won Gi Jeon ,Kyung Hi Chang and Yong Soo Cho. (1997). An Apaptive Data Predistortion for Compensation of Nonlinear Distortion in OFDM System. IEEE Transaction on communications , Vol 45 ,No .10.

[17] F. ZAVOSH, MTHOMAS, C. THRON, THALL, D. ARTUSI, D. ANDERSON, D. N. AND DAVID R. (1999). DIGITAL PREDISTORTION TECHNIQUES FOR RF POWER AMPLIFIERS WITH CDMA APPLICATIONS. TECHNICAL FEATURE, MICROWAVE JOURNAL.

[18] B. Abdulrahman, G.Baudoin. (2002). Applying Digital Predistortion To Power Amplifiers Used in Third Generation Systems. ESIEE, Signal Processing and Telecommunications Department. BP-99, 93162

[19] By Kelly M., Wan-Jong Kim, Shawn P. Stapleton, Simon Fraser University Jong Heon Kim, K. University . (2004). Linearizing Power Amplifiers Using Digital Predistortion, EDA Tools and Test Hardware. High Frequency Design, High Frequency Electronics. AMPLIFIER LINEARIZATION

[20] J. de Mingo and A. Valdovinos. (1997). Amplifier linearization using a new digital predistorter for digital mobile radio systems. IEEE 47th Vehicular Technology Conf., vol. 2, 671–75

[21] A. Mansell and A. Bateman. (1994). Practical implementation issues for adaptive predistortion transmitter linearization. IEE, London, U.K., WC2R 0BL.

[22] Kathleen J. M., Mohsen K., Fellow, IEEE, and Rajeev K. (2000). Look-Up Table Techniques for Adaptive Digital Predistortion: A Development and Comparison. IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 49, NO. 5



[23] R Singla and SK Sharma. (2012). Low complexity look up table based adaptive digital predistorter with low memory requirements. nication and Networking , EURASIP Journal on Wireless Commu Singla and Sharma EURASIP Journal on Wireless Communications and Networking


Maryam Sajedin was born in Tehran, Iran, on April 4, 1986. She received the B.Sc. degree in Electrical Engineering in 2008, She is currently working toward the M.Sc degree in communication engineering at Islamic Azad University, fars, Iran since 2014. Her working experiences are digital communication, nonlinear power amplifier and application of signal processing in multimedia communication system. her research interests include nonlinear modeling of HPA and compensation techniques for nonlinear distortion in OFDM system.

Ayaz Ghorbani received Postgraduate Diploma, M.Phil., and Ph.D. degrees in electrical and communication engineering as well as postdoctoral degree from the University of Bradford, UK, in 1984, 1985, 1987, and 2004, respectively. Since 1987 up to now he has been teaching various courses in the Department of Electrical and Electrical Engineering, AmirKabir University of Technology (Tehran Polytechnic), Tehran, Iran. Also from 2004 to 2005, he was with Bradford University for sabbatical leave. He has authored or coauthored more than 120 papers in various national and international conferences as well as refereed journals. In 1987, Dr. Ghorbani received John Roberts haw Travel Award to visit

USA. In 1990, he received the URSI Young Scientists Award at the General Assembly of URSI, Prague, Czech Republic. He also received the Seventh and Tenth Kharazmi International Festival Prize in 1993 and 1995 for designing and implementation of anti-echo chamber and microwave subsystems, respectively. His research areas include Radio wave propagation, antennas bandwidth, nonlinear modeling of HPA, antecho chambers modeling and design, electromagnetic shielding as well as EMI/EMC analysis and modeling. He has authored one book in Microwave circuit and devices.

Hamidreza Amindavar received B.Sc., M.Sc., M.Sc.AMATH, and Ph.D. degree from the University of Washington in Seattle, in 1985, 1987,and 1991, respectively, all in electrical engineering. He is currently a Professor in the Department of Electrical Engineering, Amirkabir University of Technology, Tehran, Iran. His research interests include signal and image processing, array processing, and multiuser detection. Prof. Amindavar is a member of Tau Beta Pi and Eta Kappa Nu.


E-Portfolio Assessment for Learning: TenYears Later – an Experience from an

Outcome-Based University

Authors

Abdallah Tubaishat College of Technological Innovation, Zayed University

[email protected] Abu Dhabi, UAE

Abstract

An e-assessment tool, dubbed e-portfolio can be an innovative tool that provide students with flexible opportunities to demonstrate the acquisition of skills and abilities in an outcome-based institution. An E-portfolio Assessment Management System (EAMS) has been developed and used to create, reflect, revise, and structure students’ work via digital medium. The system is a web-based e-portfolios which was developed in-house. It is a repository management system that facilitates collecting, sharing, and presenting artifacts of student learning outcomes via a digital medium. The rationale of the EAMS is to allow students to present a collection of items that represent their accomplishments in courses towards the satisfaction of pre-determined courses learning outcomes using a pedagogical web-based environment model. The system was built around two defined set of learning outcomes: institutionally agreed upon set of learning outcomes, and learning objectives that are related to major requirements. The purpose of this research is analyze students’ perceptions of using EAMS to support their learning and assessment in an outcome-based institution after ten years of implementation. The participants were 217 students in IT college. The results showed that the students had positive opinions about using e-assessment tool: It enhanced their learning through reflection; assisted them monitor their academic progress towards achieving their degree programs; helped them identify strength and weaknesses by reflecting on their work; and made assessments of artifacts more effective and efficient, hence according to students, the evaluation of students’ e-portfolios is a better way to assess students’ knowledge than using tests or exams. In conclusions, the e-assessment system has a significant and positive influence on self-perceived learning performance where students are accountable for their learning. Furthermore, our evaluation uncovered organizational, learning, and technological issues involved in moving from traditional approach of teaching learning toward an integrated learning system approach.

Key Words

E-portfolio, Assessments, Learning Curriculum, Evaluation, Student Perspectives, Outcome-Based Higher

Education.

E-Portfolio Assessment for Learning: Ten Years Later – an Experience from an Outcome-Based University Abdallah Tubaishat


I. INTRODUCTION An e-portfolio can be defined as a collection of student work accomplished throughout a

student’s time in an academic program [1]-[5]. Therefore, an e-portfolio can demonstrate growth of achieving learning outcomes to measure what students have learned and are able to do when they complete their degree. E-portfolio has been widely used in education for collecting students’ artifacts. However, it can also includes a series of reflections that allows understanding the learning and teaching process as well as facilitating the evaluation [6]. Wheeler defined an e-portfolio as “a collection of purposefully-organized artifacts that support backward and forward reflection to augment and assess growth over time” [1]. Paulson and Meyer described a portfolio as a meaningful collection of student work that demonstrates progress and/or mastery guided by standards and includes evidence of student self-reflection [2]. Buzzetto indicated that electronic portfolios are a unique way to document student progress, encourage improvement and motivate involvement in learning [3]. Buzzetto [3] and Wright [7] defined portfolios as an effective form of alternative assessment that encourages students and educators to examine skills that may not be otherwise accessed using traditional means such as higher order thinking, communications, and collaborative abilities. Miller and Morgaine [8] noted that “E-portfolios provide a rich resource for both students and faculty to learn about achievement of important outcomes over time, make connections among disparate parts of the curriculum, gain insights leading to improvement, and develop identities as learners or as facilitators of learning”. Chambers and Wickersham [9] indicated that more and more institutions are moving to use e-portfolio in education because these “institutions are expected to be more accountable for providing evidence of the process and growth in student learning during their academic progress”. Furthermore, “student learning outcomes have become the focus of many universities as a way to measure and document student learning. These outcomes measure how a student's university experience has supported their development as individuals and describes the knowledge, skills, abilities and attitudes students are able to demonstrate upon completion of a program. A learning outcome is not what the instructor does to the student, but rather what the instructor expects the student to do as a result of teaching.” Chambers and Wickersham [9] noted that “the assessment of student learning outcomes within one course is not a new concept; however, tracking and assessing student learning and providing data for learning outcomes for assessment of learning and for learning within an entire program presents a new challenge.”

From the literature survey we conclude that e-portfolio can provide rich resource for both students and faculty to learn about accomplishment of important learning outcomes over years of study. Hence, the process of building e-portfolio in educational institutions encourages the establishment of clear learning goals and expectations. According to Lorenzo and Ittelson [10] e-portfolios are the biggest innovation in educational technology since the introduction of course management systems such as Blackboard and WebCT. Gulbahar & Tinmaz [11] indicated that e-portfolio is one of the newest evaluation techniques for new learning environments in which students show their artifacts, products and projects as an indication of their functional learning. E-portfolios are being integrated relatively quickly into colleges. Literature illustrates a number of organizational purposes, including meeting requirements from accrediting boards and states'



www.elvedit.com 370

approved technology standards while offering students an alternate form of assessment [12]. Early pioneers in implementing successful digital portfolios are Alverno College, University of Minnesota, and Indiana University-Purdue University Indianapolis (IUPUI).

We view e-portfolio as an innovative assessment tool that provide students with flexible opportunities to demonstrate the acquisition of skills and abilities. It is used to plan, collect, and store multiple sources of data maintained in the portfolio. Therefore, it is being recognized as being a technological tool that allows the student to manage their learning experience. E-portfolio is based on a systematic assessment procedure that can provide accurate information about the depth and breadth of a student's capabilities in many domains of learning. The system recognizes the following three levels of assessments:

- Student Level: Creating a system of tracking student work over time, in a single course, with students and faculty reflecting on it.

- Course Level: Aggregating many students' work in a particular course to see how the students as a whole are progressing toward learning goals.

- Program level: Assessing many courses in similar ways that are all part of one major and thus, by extension, assessing the entire program of study.

II. PURPOSE OF THE STUDY An e-assessment tool has been developed and used at an outcome-based institution for the

past ten years. However, no comprehensive study has been conducted to evaluate and asses the deployment impact on student learning. The main objective of this research is to investigate how students perceive the e-assessment tool in IT College and if it made academic assessment activities more effective and efficient. We also anticipate that as an accredited institution by middle state (MSHE) as well as other degree programs from internationally accredited bodies such as ABET, the evaluation of the system will optimize the process for accreditation.

III. THE INSTITUTION UNDER STUDY Zayed University (ZU) is an academic public institution in United Arab Emirates (UAE). It

offers an academic program that prepares students for success in education, arts, business, media and IT in two campuses one in Abu Dhabi and one in Dubai. ZU is concerned with “outcome assessments”, how learning and growth are measured, evaluated, and demonstrated over years of study. Today, the University is educating more than 9,000 male and female students. The university endeavors to provide students learning opportunities using the American style of education and learning to ensure high quality education. The University is accredited by the Middle States Commission on Higher Education (MSCHE) in 2008. Majority of the faculty members are from North America, Europe, or Australia, or have a terminal degree from universities from developed countries [14].

ZU has an excellent technology infrastructure; its campuses are fully networked and allow students to connect to various university networks and the Internet from anywhere on campus. All the university has wired and wireless connections (classrooms, library, offices, student hubs,



cafeteria, etc). Each student is required to purchase a laptop and each faculty member receives a laptop with a three-year replacement schedule. Students have easy access to technology in order to facilitate the learning process. Actually, ZU is known as the laptop university in this region. In the College of Technological Innovation (CTI), students have their own laptop loaded with the necessary software for their courses. This helps them complete their work independently without having to be on campus all the time. The CTI College has an independent network infrastructure for teaching and research, in addition to the university’s main network. This infrastructure allows students to login remotely into Linux servers to use tools needed for programming languages, databases, and web development courses. Students can also use Linux-based communication tools to collaborate with each other and with instructors. All ZU courses are implemented on Blackboard Learn+, a learning management system. ZU students can access Blackboard Learn+ from anywhere at any time using a web client.

ZU has adopted an outcome based learning framework to provide a strong focus to the students’ learning outcomes and to improve both curriculum and learning practices. The Academic Program Model (APM) was developed by faculty and emphasizes a commitment to a learner-based education and to shift the teaching paradigm to a student learning model. This model focuses on what students can actually do after they graduate. More details about this model can be found in the ZU internal report on “Self-Assessment Based on Accreditation Standards of the Middle States Commission on Higher Education”, and the ZU Academic Program Model [14]. The purpose of the outcome-based model is to provide students with a focused and coherent academic program and to prepare graduates for a rapidly changing and unpredictable future. It is outcome driven and uses the traditional Grade Point Average (GPA) system. The framework that constitutes the academic program model is composed of three components:

- Readiness program to ensure that students are competent in English language - General Education - Degree Major

A major objective of the undergraduate experience at ZU is the development of the skills

necessary for continuous lifelong learning. The APM is designed to help achieve this objective by providing students with a foundation and framework for all university studies. Every ZU course focuses on one or more of the six university-specified learning outcomes. The learning outcomes are incorporated into normal course work and, therefore, are an integral part of disciplinary content and evaluation of the course. Threaded throughout the baccalaureate curriculum, the learning outcomes help students achieve a higher order of intellectual development. ZU has six graduation requirements, called Zayed University Learning Outcomes (ZULOs), for all students regardless of their major. These requirements are depicted in Table 1 [14].



www.elvedit.com 372

TABLE 1: ZAYED UNIVERSITY LEARNING OUTCOMES

Learning Outcome Description Information Literacy and Communication

ZU graduates will be able to recognize information needs, access and evaluate appropriate information to answer those needs, and communicate effectively to a variety of audiences in both English and Arabic.

Information Technology ZU graduates will be critically aware of the implications of information technology on the individual and on society and be able to use IT to communicate and solve problems in an ethical way.

Critical Thinking and Quantitative Reasoning

ZU graduates will be able to use information, reasoning, and creative processes to achieve goals and make responsible decisions.

Global Awareness ZU graduates will be able to relate to communities beyond the local, perceive and react to differences from an informal and reasoned point of view, and be critically aware of the implications and benefits of cultural interaction.

Teamwork and leadership

ZU graduates will be able to work efficiently and effectively in a group. ZU graduates will be able to assume leadership roles and responsibilities in a variety of life situations and accept accountability for the results.

Bilingual ZU graduates will be able to communicate effectively (orally and in writing) in both English and Arabic.

IV. OUTCOME-BASED CURRICULUM Student learning outcomes have become the focus of many universities as a way to measure

and document student learning [9]. Chambers and Wickersham indicated that “these outcomes measure how a student's university experience has supported their development as individuals and describes the knowledge, skills, abilities and attitudes students are able to demonstrate upon completion of a program”. Furthermore, the methods by which these learning outcomes are assessed to determine student success of learning expectations vary and may be dependent upon the course, program, and/or assessment practices and beliefs of the faculty.

The Information Technology program under study strives to meet the demands of government and industry in the UAE technology market. This cooperative process usually includes advisory boards, called National Advisory Council (NAC), where industry leaders communicate the technical needs to faculty and administrators. Currently, the CTI College offers four tracks: Security and Networking, Enterprise Computing, Multimedia Design, and Business Information System. All core courses in each sequence include specific university learning outcomes (ZULOs) and specific major learning outcomes (MALOs) that are applicable to the courses contents. The CTI College has established six learning outcomes that complement the learning outcomes of the ZU APM (see Table 2). These major learning outcomes form the basis for analysis and assessment that play an essential role in the continuous process of improvement.

TABLE 2: MAJOR LEARNING OUTCOMES FOR CTI COLLEGE Learning Outcome Description Critical Thinking and Quantitative Reasoning in IT

CTI College graduates will be able to use critical thinking and quantitative processes to identify, analyze and solve problems, and evaluate solutions in an IT context.

Information Technology Application

CTI College graduates will be able to select existing and cutting-edge IT tools and procedures to develop modules and systems.



Learning Outcome Description Information Technology Management

CTI College graduates will be able to assess and determine information resource requirements to develop solutions suitable for IT and business managers operating in a multi-national and multi-cultural environment.

Information Technology Professional Practice

CTI College graduates will be able to work effectively in individual and group situations, understand how groups interact, be able to assume a leadership role when required, and understand the fundamentals of professional and ethical conduct.

Information Technology Systems Theory and Practice

CTI College graduates will be able to understand and communicate the fundamentals of systems theory in the development of appropriate systems that function in a global environment.

Technical Communication (Bilingual)

CTI College graduates will be able to express themselves effectively and efficiently in both English and Arabic while using the correct IT terms for each language.

V. E-PORTFOLIO ASSESSMENT FOR LEARNING The EAMS is an important resource used in the colleges for various assessment activities. The

system is a searchable, electronic storage tool into which specific examples of student work are uploaded from various courses across the curriculum. Students regardless of their major start using EAMS in semester three of their degree programs and therefore, they begin the development of a working e-portfolio by archiving assignments, instructors’ feedback and reflections during early courses.

In each course in the Information Technology concentrations, faculty members are required to assign assignments designed to assess at most two of the six MALOs presented in Table 2. Because the e-portfolio assignments are the key to the success of the outcome assessment process, faculty members are encouraged to design assignments that provide students with an opportunity to demonstrate their most distinguished performance and scholarly accomplishments. Examples of appropriate e-portfolio assignments include a term paper, a project work, a programming assignment, or a network design. Faculty members are required to provide a criteria sheet for each portfolio assignment that explains the purposes and the learning objectives assessed. After reviewing student work, the faculty comments on the student’s work and posts their feedback in a designated area of the EAMS. Moreover, the faculty evaluates both the assignment’s general effectiveness and its level of accomplishment with respect to the desired outcome(s). Students are able to access the faculty comments from the EAMS, as well as any other work posted on the system. This process enables students to update their work and reflect on their learning. The EAMS was designed to function as an archive for research on the effectiveness of various courses in achieving learning outcomes. Because all major student work is uploaded to the system, research into student achievement of learning outcomes in courses or sequences of courses can be easily carried out. Furthermore, student work can be sorted and studied either by course or by outcome. Figure 1 shows the EAMS interface. The interface shows for a particular faculty member the courses being taught as well as the assessment criteria posted by the instructor. It allows him/her to select term code and the courses in that semester through accessing e-portfolio systems via the Intranet or extranet.



www.elvedit.com 374

FIGURE 1: E-PORTFOLIO ASSESSMENT MANAGEMENT SYSTEM INTERFACE

After matching courses with learning outcomes, faculty members develop key assignments for

the courses to optimize targeted learning. The assignments include a term paper, a lab exercise, a design for building a network, or a case study. Figure 2 shows another snapshot of the EAMS for an e-portfolio course (CIT490) with committee member names, learning outcomes, and assessment criteria used in this course. After grading the piece of evidence, the instructor posts the assessment feedback. The students can then access the instructor’s assessment/feedback and modify their work. Finally, the students have the option to include that piece of evidence as an artifact in their e-portfolio.

It is to be noted that the CTI College has accumulated a significant amount of data from the EAMS to evaluate students’ achievements. Figure 2 shows another snapshot of the EAMS for a Course Review Report. In this view, we can generate specific reports about semester, course(s), and type of assignments (e.g. Project, Term Paper, presentation, etc.).



FIGURE 2: E-PORTFOLIO COURSE REVIEW REPORT

VI. CASE STUDY The purpose of the study is analyze students’ perceptions of using EAMS to support their

learning and assessment in an outcome-based institution after ten years of implementation. A survey is conducted in CTI College to identify and analyze students’ opinion of the e-assessment tool to support their learning; and to recognize if the e-assessment tool has made academic assessment activities more effective and efficient. The specific research questions that guided the study are shown below (Table 3).

TABLE 3: RESEARCH QUESTIONS

A. Participants

The participants of the study were volunteered randomly from the baccalaureate program majoring in Information Technology from all levels. The institution is a public outcome-based university located in the Gulf region. Data was collected from 217 students, 145 female, and 72 male. Data is collected in Srping 2014 through an online questionnaire. The survey was distributed online. The questionnaire was made up of 7 closed-ended questions with five multiple choice questions (Strongly Disagree, Disagree, Undecided, Agree, Strongly Agree).

Does e-portfolio allow students to get better feedback on their work? Does e-portfolio help students to reflect on their work? Does e-portfolio help students monitor their progress towards achieving goals in their degree program? Does e-portfolio give students a chance to provide constructive feedback to their and other peers work? Does e-portfolio help students use feedback from their teachers to improve their work? Is the evaluation of e-portfolio a better way to assess students’ knowledge than using tests or exams? Is it a good idea to use e-portfolio as part of the evaluation process to capstone (final project) course?



www.elvedit.com 376

B. Survey Analysis

A survey questionnaire was developed to gather data to try to find an answer to the above questions and also learn about the student’s attitudes and opinions about using the e-assessment tool: Seven questions were developed to gather the data needed to answer the research questions. Table 4 shows all the questions with the respective students’ responses and analysis.

TABLE 4: QUESTIONNAIRE RESPONSE ANALYSIS

My e-portfolio allows me to get better feedback on my work: The first question was designed to learn about the student’s opinion on whether they received better instructor feedback using e-assessment tool. Around 75 percent responded with agreement (or strongly agreed) to a question regarding that the use of the e-assessment tool helps them get better feedback from the instructors. Around 16 percent of the students were undecided about that perception and only 9 percent disagreed or strongly disagreed about the fact that e-assessment tool allows them to get better feedback from their instructors. My e-portfolio helps me to reflect on my work: The second question was designed to learn about the possibility that e-assessment tool can help students in their reflections after receiving the instructor’s feedback on their posted work. More than two third of the students responded by either agreed or strongly agreed to the fact that the use of their e-portfolio helped them in reflecting on their work. About 15 percent were undecided and 9 percent either disagreed or strongly disagreed that using their e-portfolios helped them reflect on their work. My e-portfolio helps me monitor my progress towards achieving goals in my degree program: This question was designed to learn about the student’s impression from the use of the e-assessment tool to help them monitor their progress towards achieving their learning goals in their degree program. About 65 percent of the students either agreed or strongly agreed to that statement, 25 percent were undecided whether their e-assessment tool helped them monitor their progress towards achieve their goals. Around 10 percent disagreed or strongly disagree to that statement. My e-portfolio gives me a chance to provide constructive feedback to my and other peers work: This question was designed to learn about the possibility that the e-assessment tool helped students provide constructive feedback to their learning experiences. These experiences usually are the student’s key pieces of evidence towards achieving learning outcomes. About 67 percent agreed or strongly agreed to that statement, about 18 percent were unsure or undecided and about 15 percent either disagreed or strongly disagreed to the fact that their e-assessment tool could help towards that goal. My e-portfolio helps me use feedback from my teachers to improve my work: This question was designed to learn about the possibility that the e-assessment tool helped students receive instructor feedback to plan and/or improve their learning experiences. Around 69 percent responded with agreement (or strongly agreed) to that statement, about 18 percent were unsure or undecided and about 13 percent either disagreed or strongly disagreed to the fact that their e-assessment could help towards that goal. I feel that the evaluation of my e-portfolio is a better way to assess my knowledge than using tests or exams: This question was designed to learn about the student’s impression regarding what is the best way for them to be evaluated . About 67 percent of the students either agreed or strongly agreed with the statement, while only 14 percent disagreed with, and around 19 percent were undecided. I like the idea of using my e-portfolio as part of a capstone (final project) course: The last question was designed to learn about the student’s opinion on whether they would prefer their collective artifacts from courses to be used as part of the final graduation project. More than 75 percent of the students either strongly agreed or agreed that they prefer their e-portfolio to be used to judge their accomplishment in the degree program. Around 16 percent of the students were undecided about that perception and only 9 percent disagreed or strongly disagreed about the fact that it is not a good idea to use their e-portfolio as part of the capstone project.



VII. CONCLUSIONS

An E-portfolio Assessment Management System (EAMS) has been developed and used in an outcome-based university to create, reflect, revise, and structure students’ work via digital medium. The purpose of this study is analyze students’ perceptions in IT College of using EASM to support their learning and assessment in an outcome-based institution after ten years of implementation. The results showed that the students had positive opinions about using EASM: It enhanced their learning through reflection; assisted them monitor their academic progress towards achieving their degree programs; helped them identify strength and weaknesses by reflecting on their work, and made assessments of artifacts more effective and efficient, hence, they think the evaluation of their e-portfolios is a better way to assess their knowledge than using tests or exams.

VIII. FUTURE ENHANCEMENT

Over ten years, our institution has identified, researched, and developed an e-portfolio assessment tool that meets the needs of the students. The tool is easy to use. Regarding our research questions and statistics, we can confirm that students in CTI College have positive attitudes towards using the e-assessment system, they developed positive attitude towards e-portfolio processes such as reflections. Currently, e-portfolio can only be used by current students. A future goal is to allow students to keep their e-portfolio after graduation. Another following step in this research would be to gather data from faculty to see if these results are in line with students’ opinion.

REFERENCES [1] Wheeler, B. C.(2014), E-Portfolio Project, Open Source e-Portfolio Release, Andrew W. Mellon

Foundation, Version 2.0, Retrieved from: http://juicy.mellon.org/RIT/MellonOSProjects/%20e-Portfolio/Portfolio_Proposal_Public.doc.

[2] Paulson, F. L., Paulson, P. R., & Meyer, C.(1991). What makes a portfolio a portfolio? Educational Leadership, 48(5), 60-63.

[3] Buzzetto-More, N. (2006). The e-Learning and business education paradigm: Enhancing education, assessment, and accountability. Proceedings of the Maryland Business Education Association Conference. Ocean City, MD.

[4] Love, D., McKean, G., Gathercoal, P. 2004). Portfolios to Webfolios and Beyond: Levels of Maturation, EDUCAUSE Quarterly, 27, 2, 24 – 37.

[5] Siemens, G., e-Portfolios (2004). eLearnSpace: Everything ELearning, Retrieved from: http://www.elearnspace.org/Articles/e-Portfolios.htm.

[6] Amaya, P., Agudo, J., Samches, H., Rico, M., Hernandez-Linare (2013). Educational e-portfolio Uses and Tools, Social and Behavioral Sciences, 93(2013) 1169 – 1173.

[7] Wright, B. (2004). An assessment planning primer: Getting started at your own institution. 13th Annual Northeast Regional Teaching Workshop, October 1.

[8] Millar, R., and Morgane, W. (2009). The Benefits of E-Portfolios for Students and Faculty in Their Own Words. AAC&U, pp. 8-12, Winter.



www.elvedit.com 378

[9] Chambers, S. and Wickersham, L. (2007). The Electronic Portfolio Journey: A Year Later. Education Journal, Vol. 127, No. 3, pp. 351-360.

[10] Lorenzo, G., & Ittelson, J. (2014). An Overview of e-Portfolios. Retrieved from http://www.case.edu/artsci/cosi/cspl/documents/eportfolio-Educausedocument.pdf.

[11] Gulbahar & Tinmaz (2006). Implementing Project-Based Learning And E-Portfolio Assessment In an Undergraduate Course, Journal of Research on Technology in Education, Volume 38, Issue 3.

[12] Ritzhaupt, A., Singh, O., Seyferth, M., Dedrick, T. (2008). Development of the Electronic Portfolio Student Perspective Instrument: An ePortfolio integration initiative, Journal of Computing in Higher Education, Volume 19, Issue 2, pp 47-71, Spring.

[13] Moya, S, O’Malley, J. (2009). A Portfolio Assessment Model for ESL. The Journal of Educational Issues of Language Minority Students, Vol. 13, pp. 13-36.

[14] Zayed University. Retrieved from http://www.zu.ac.ae.


Abdallah Tubaishat is an Associate Professor in the College of Technological Innovation at Zayed University, United Arab Emirates. He received his PhD in Software Engineering from Illinois Institute of Technology, IL, USA in 1994. Dr. Tubaishat has twenty years of experience in teaching and research. His teaching experience include: Software Engineering, Database and Programming, His research spans two main areas, one is technical: software engineering, and the other is non-technical: e-learning, and

educational technology. He has published a book with others entitled "Computer Skills", and has around twenty three Journal and conference publications. Dr. Tubaishat served on the program and organizing committees of several international conferences and workshops.


How Programmer Plans Training?

Authors

Jakub Novotný Department of Electrical Engineering and Computer Science, College of Polytechnics Jihlava

[email protected] Jihlava, Czech Republic

Martina Winklerová Department of Languages, Department of Sports, College of Polytechnics Jihlava

[email protected] Jihlava, Czech Republic

Abstract

The paper describes the newest trend of small downloadable applications for high-end GPS sport watches. Basics facts from history of sport watches with monitoring of heart rate and basic facts about use of heart rate in sport training are mentioned. The concept of applications is demonstrated on the code of three Suunto Apps created by the authors of this paper. First demonstrated app called MINIMUM HR LOCATOR finds minimum hearth rate during measurement. Second app called FIND MAXIMAL HR - ENDURANCE RUNNERS guides athletes through exercise to find maximal heart rate. Last demonstrated app called FARTLEK TRAINING guides athletes through fartlek training controlled by current heart rate. All three apps were tested on appropriate device and published in App zone on portal moverscount.com. The paper concludes that there are limitations in simplicity and small range of Suunto Apps scripting language and the solution is highly proprietary only for one product line of devices. But in good combination of programming and sport training knowledge it can result in very effective extension in functionality of the sport watches.

Key Words

Sport watch, heart rate, applications, Suunto Apps, sport training.

I. INTRODUCTIONOn September 24th (2014) Garmin announced open platform called Connect IQ for third-party

developers to create applications for Garmin wearable products [1]. Among other aims, it is certainly a response to concept of downloadable applications called Suunto Apps which introduced Suunto in early 2013. Downloadable applications (apps) seem to be “a must” in high-end GPS sport watches. We can classify these types of equipment generally as mobile technical



www.elvedit.com 380

means of personal information and communication technologies (ICT).

According to document Market Developments – Remote Patient Monitoring and Treatment, Telecare, Fitness/Wellness and mHealth [2] the development of sport and fitness market sector of personal ICT is very significant for future development of broader segment of ICT for personal healthcare and the fastest development in segment of ICT means for sport and fitness is focused on general public and not on sport professionals. Market leaders in this sector are well aware of this potential and so a wide variety of products and solutions directed to this area can be found - from various hardware accessories to sport and fitness software applications added to the products designed primarily for another type of use (smartphones etc.). On the other hand completely designed equipment is being developed and offered in the market.

Sport watches with the function of hearth rate monitor have the historically longest period in the market and they are also probably the most widely used device.

It is usually a sport watch displaying and recording the actual heart rate of their users most often via wirelessly connected chest belt. This concept including the sometimes used generalized name "sporttester" for a similar type of device comes from the Finnish company POLAR. According to information presented by POLAR the idea originated in 1975 during a cross-country skiing tour of future company founder and national coach of the Finnish cross-country ski team. Then in 1977 at the University of Oulu a finger pulse sensor was developed and its market variant was introduced in 1978 by POLAR. A prototype of wireless heart rate monitor was introduced at the same time, initially designed for the needs of the Finnish cross-country ski team. Shau Parker states: “The wireless Polar heart rate monitoring method was developed at the University of Oulu's department of electronics, and was originally aimed at coaches and sportsmen to help raise the quality and efficiency of their training. Exercise scientists also used them in their work after researching them [3].” With the gradual expansion of methods of training by heart rate and also with the market availability of appropriate equipment to monitor heart rate it began to be used also in the amateur and recreational sports. In terms of "market access" year 1982 was a major breakthrough (according to some sources 1983), when company POLAR began offering device Sport Tester PE2000 – “The world's first wireless wearable heart rate monitor [4].” Other models also allow the evaluation of the training load and its effect through connection to a computer (IBM PC). From the point of view of today's boom of various smart mobile devices and means like “smart” watches we can with no exaggeration say that the heart rate monitors were ahead of these trends by more than a quarter of a century.

Current offer of "smart" sports watch with heart rate function is very broad. Almost all high-end models use GPS signals to specify distance measurements during athletic activity or even data from other internal or external sensors. The main leaders in this segment are undoubtedly companies (in alphabetical order) GARMIN, POLAR and SUUNTO.

In this paper we focus in more details on the use of Suunto Apps (as a first market solution of downloadable apps for sport watches) in sport training by the Finnish manufacturer Suunto. It should be demonstration of development of "smart" sports watch since 1982 and mainly demonstration of newly expanded concept of apps in electronic devices.

How Programmer Plans Training? Jakub Novotný and Martina Winklerová


II. HEART RATE AS AN INDICATOR IN SPORT TRAINING Why to monitor heart rate at all? The registration of heart rate belongs among the most

frequently used methods of examinations in physical exercises and sports. Heart rate indicates present state of human body and its ability to adapt to stress provoked by physical activity. The absolute heart rate value is the basic indicator in the evaluation of human adaptability to various intense physical activity and crucial value for setting intensity intervals of sport trainings. Together with the evaluation of heart rate variability observed mostly before and after the load and the heart rate measured during the load, the absolute heart rate value give us the basic data for evaluation complex view of the behavior and adaptability of the organism in the course of sport training.

Rest heart rate (HRrest) and Absolute heart rate (HRmax) are two basic parameters. HRrest expresses beats of heart per minute during the rest regime. HRrest is under training influence and it usually decreases with growing performance. The best time to measure minimum heart frequency (HRmin) is in the morning after waking up. Sometimes it can grow which can indicate fatigue, illness, stress or overtraining.

HRrest and HRexe (exercise) are influenced with several factors. Fitness and the level of recovery are among the most important. The fitter people have usually lower HRrest. The suitably set training method makes heart stronger and bigger. The stronger heart has bigger cardiac volume and can do the same work with fewer beats. It means that HRrest should get lower during training process.

The human body needs recovery period after each training. The muscles are damaged, energy supplies are exhausted. The HRrest is higher because of recovery processes that are in motion. If the next training follows too soon after the previous one, the organism hasn´t recovered yet and in this case the HRexe can be higher and training less effective.

HRmax shows the maximum heart beats per minute, it is strictly individual and stable (it doesn´t change with growing performance) and all training zones should be set according to its value. That is why the knowledge of HRmax is crucial information for correct training plans. There are two main procedures how to get HRmax:

A. Calculations B. Load tests

A few formulas of counting HRrest are used [5], [6]:

1. HRmax (men) = 202 – (0,55 . age) HRmax (women) = 216 – (1,09 . age)

2. HRmax (men) = - 0,4635 . age + 202 HRmax (women) = - 0,5148 . age + 206

3. HRmax = 220 – (0,5 . age)

4. Karvonen formula



www.elvedit.com 382

HRmax = 220 – age HRrez = HRmax = HRrest Intensity of the training: % . HRrez + HRrest

HRmax values at population can be displayed with using bell curve. The most of people belong to the wide range from one end of the curve to the other. These formulas can be used by those individuals, who gains expected values. Still there are some differences in results coming from above mentioned ways of counting. Example for 40 years old man:

1. HRmax = 202 – (0,55 . 40) = 180 beats per min 180 . 0,7 (70%) = 126 beats per min 180 . 0,7 (80%) = 144 beats per min Training intensity range: 126 – 144 beats per min

2. HRmax = - 0,4635 . 40 + 202 = 183 beats per min 183 . 0,7 = 128 beats per min 183 . 0,8 = 146 beats per min Training intensity range: 128 – 146 beats per min.

3. HRmax = 220 – 40 = 180 beats per min 180 – 70 (HRrest) = 110 beats per min (HRrez) 0,7 . 110 + 70 = 147 beats per min 0,8 . 110 + 70 = 158 beats per min Training intensity range: 147 – 158 beats per min

As we can see, using different formulas can bring different results. The calculations besides don´t take into consideration other physical factors that can have impact on HRmax, for example the size of heart. This is the reason why load tests should be used to set HRmax rather than calculations.

These tests can be done by professionals in laboratory conditions using telemetric system observing HR, power output, speed, time etc. The recent researches show that HRmax can differ in dependence on concrete sport activity during which it is measured. It means that runners should set their HR max on running simulator, cyclists on cycling simulator etc. [9] But even these tests can bring wrong results if they aren´t taken well and are finished prematurely. The real HFmax is the value which we have at the moment when it is stable and doesn´t change no matter how long or with which intensity the physical activity lasts. It is usually the state of physical exhaustion and only few sportsmen are able to meet this crucial point when taking part in the test.

Benson and Connolly introduced an easy test that can help amateur sportsmen to find their HRmax. [7]

1. Find running track of 400 meters. 2. Take the sport tester. 3. The 1st lap – walk at normal speed. 4. The 2nd lap – walk at brisk speed. 5. The 3rd lap – jog at the slowest possible speed. 6. The 4th lap – jog at quiet conversational pace.



7. The 5th lap – faster running, still it is possible to talk, light breathing. 8. The 6th lap – fast running, it isn´t possible to talk. 9. The 7th and 8th lap – to increase speed and effort at the beginning of each turn. The last

half of the last lap should be run at high speed.

Monitor HR each 200 meters. In the course of last two laps monitor HR four times. The highest value of HR monitored during or immediately after the test should equal the HRmax.

III. SUUNTO APPS As mentioned above the first concept of downloadable specific (small) applications was

introduces by Suunto. Suunto Apps allow also to create own applications and add them to the sport modes. Proprietary scripting language is used (based on Java script obviously) for writing of these applications. Publicly downloadable applications are then available in the App Zone of portal movescount.com. A tool called App designer is available on the same portal for the creation of applications. App designer is working in two user profiles - simplified (graphic creation mode of applications) and advanced one. The advanced version is a full-fledged writing of the script and its validation with the possibility of testing prior to uploading it to the device. There is available reference Guide called Suunto Apps - Developer manual for advanced creation of Suunto Apps [8].

The basic output of the Suunto script is figure (RESULT) displayed in time format or numbers. This output can be supplemented by the text field before and after the output (PREFIX, POSTFIX). Total number of displayable characters together for PREFIX and POSTFIX is only six. When creating Suunto Apps basic mathematical and logical operators, some mathematical functions and so called Suunto functions are available (beep, backlight, distance and heading between two points according to the geographical coordinates). Over 200 values (variables) are also recorded by the device such as the speed characteristics (current speed, average speed, maximum speed, pace, etc.), distance characteristic, heart rate, ambient and many others.

There is number of limitations and difficulties when writing Suunto App Script such as very small memory available for application, considerable constraints of scripting language or the limited display capabilities. However, in the words of one portal movescount.com user: "It's challenging to write for dry small memory footprint something strong [9]." Introduced competitive concept Connect IQ [10] for Garmin with a complex object-oriented programming language Monkey C should be much more powerful, but some hardware limitations will still prevail.

IV. HEART RATE DRIVEN TRAINING WITH USE OF SUUNTO APPS In the following paragraph we would like to demonstrate some specific solution resulting from

above part Heart rate as an indicator in sport training by Suunto Apps. First demonstration is code of Suunto application helping to find and display HRmin. The first version of this app displayed just only absolute HRmin during measured period, but by other testing the authors discovered that initial data from sensor (chest belt) often contained also zero values and it distorted measurement of HRmin. So HRmin is evaluated in 120 seconds in the second version and moreover as moving average for 30 seconds. In another version, therefore the minimum value is evaluated in two minutes as an average of 30 s. Futhermore values under 25 beats per minute



www.elvedit.com 384

are not evaluated. Code of resulting apps has the form:

if (SUUNTO_DURATION < 120) {RESULT = SUUNTO_HR; prefix = "HR";} if (SUUNTO_DURATION >= 120) {if (MinHR <= SUUNTO_HR_AVG[30]) {RESULT = MinHR; prefix = "MinHR";} if (MinHR > SUUNTO_HR_AVG[30] && SUUNTO_HR_AVG[30] < 25) {RESULT = MinHR; prefix = "MinHR";} if (MinHR > SUUNTO_HR_AVG[30] && SUUNTO_HR_AVG[30] >= 25) {MinHR = SUUNTO_HR_AVG[30]; RESULT = MinHR; prefix = "MinHR";}}

FIGURE I: PREVIEW OF THE APP 1 IN EMULATOR ON MOVESCOUNT.COM.

Second demonstration is the app leading athlete (endurance sports) through Benson and Connolly test (mentioned above) for finding HRmax. The exercise should be done on athletic track or some slight hill road 400 - 600 m long. But there is no need to do it specifically on athletic track, because the app counts distance by GPS distance. The code of this app separately staggered is presented below:

Before athlete starts training in the relevant sport mode of watch the display shows instruction "STAR0T" (due to the fact that result must be numeric and text is limited to 6 characters)

if (SUUNTO_LAP_NUMBER == 0) {RESULT = 0; prefix = "STAR"; postfix = "T";}

Than instruction “WALK” is displayed on watch and the remaining number of meters required to 400 m after pressing button START. After reaching 400 m there is a beep and displayed instruction “LAP 2 Go”

if (SUUNTO_LAP_NUMBER == 1 && SUUNTO_LAP_DISTANCE < 0.4 ) {RESULT = (0.4 - SUUNTO_LAP_DISTANCE)*1000; prefix = "WALK"; postfix = "m";} if (SUUNTO_LAP_NUMBER == 1 && SUUNTO_LAP_DISTANCE >= 0.4 ) {RESULT = 2; prefix = "LAP"; postfix = "Go"; Suunto.alarmBeep();}

In the same manner there are instructions for following laps: WLK2 – faster walk



JOG - jog JOG2 – faster jog RUN - running RUN2 – faster running DASH - sprint DSH2 - sprint at your maximal speed

The needed code of app is than:

if (SUUNTO_LAP_NUMBER == 2 && SUUNTO_LAP_DISTANCE <= 0.4) {RESULT = (0.4 - SUUNTO_LAP_DISTANCE)*1000; prefix = "WLK2"; postfix = "m";} if (SUUNTO_LAP_NUMBER == 2 && SUUNTO_LAP_DISTANCE > 0.4) {RESULT = 3; prefix = "LAP"; postfix = "Go"; Suunto.alarmBeep();}

if (SUUNTO_LAP_NUMBER == 3 && SUUNTO_LAP_DISTANCE <= 0.4) {RESULT = (0.4 - SUUNTO_LAP_DISTANCE)*1000; prefix = "JOG"; postfix = "m";} if (SUUNTO_LAP_NUMBER == 3 && SUUNTO_LAP_DISTANCE > 0.4) {RESULT = 4; prefix = "LAP"; postfix = "Go"; Suunto.alarmBeep();}

if (SUUNTO_LAP_NUMBER == 4 && SUUNTO_LAP_DISTANCE <= 0.4) {RESULT = (0.4 - SUUNTO_LAP_DISTANCE)*1000; prefix = "JOG2"; postfix = "m";} if (SUUNTO_LAP_NUMBER == 4 && SUUNTO_LAP_DISTANCE > 0.4) {RESULT = 5; prefix = "LAP"; postfix = "Go"; Suunto.alarmBeep();}

if (SUUNTO_LAP_NUMBER == 5 && SUUNTO_LAP_DISTANCE <= 0.4) {RESULT = (0.4 - SUUNTO_LAP_DISTANCE)*1000; prefix = "RUN"; postfix = "m";} if (SUUNTO_LAP_NUMBER == 5 && SUUNTO_LAP_DISTANCE > 0.4) {RESULT = 6; prefix = "LAP"; postfix = "Go"; Suunto.alarmBeep();}

if (SUUNTO_LAP_NUMBER == 6 && SUUNTO_LAP_DISTANCE <= 0.4) {RESULT = (0.4 - SUUNTO_LAP_DISTANCE)*1000; prefix = "RUN2"; postfix = "m";} if (SUUNTO_LAP_NUMBER == 6 && SUUNTO_LAP_DISTANCE > 0.4) {RESULT = 7; prefix = "LAP"; postfix = "Go"; Suunto.alarmBeep();}

if (SUUNTO_LAP_NUMBER == 7 && SUUNTO_LAP_DISTANCE <= 0.4) {RESULT = (0.4 - SUUNTO_LAP_DISTANCE)*1000; prefix = "DASH"; postfix = "m";} if (SUUNTO_LAP_NUMBER == 7 && SUUNTO_LAP_DISTANCE > 0.4) {RESULT = 8; prefix = "LAP"; postfix = "Go"; Suunto.alarmBeep();}

if (SUUNTO_LAP_NUMBER == 8 && SUUNTO_LAP_DISTANCE <= 0.4) {RESULT = (0.4 - SUUNTO_LAP_DISTANCE)*1000; prefix = "DSH2"; postfix = "m";} if (SUUNTO_LAP_NUMBER == 8 && SUUNTO_LAP_DISTANCE > 0.4) {RESULT = 9; prefix = "LAP"; postfix = "Go"; Suunto.alarmBeep();}

Last (9th) lap is than for calming with instruction “JOG” and calm down for 2 minutes. After reaching 2 minutes for calming HRmax measured during the exercise and instruction to end whole exercise “STOP!” is displayed in 3 seconds intervals. This command is then shown even in case of another pressing of LAP button by user:



www.elvedit.com 386

if (SUUNTO_LAP_NUMBER == 9 && SUUNTO_LAP_DURATION <= 120) {RESULT = (120 - SUUNTO_LAP_DURATION); prefix = "JOG"; postfix = "s";} if (SUUNTO_LAP_NUMBER == 9 && SUUNTO_LAP_DURATION > 120) {if (Suunto.mod(SUUNTO_TIME, 5) <= 2) {RESULT = SUUNTO_MAX_HR; prefix = "Max";

postfix = "HR";} if (Suunto.mod(SUUNTO_TIME, 5) > 2) {RESULT = 0; prefix = "STOP"; postfix = "!"; Suunto.alarmBeep();}} if (SUUNTO_LAP_NUMBER > 9){RESULT = 0; prefix = "STOP"; Suunto.alarmBeep();}

FIGURE II: PREVIEW OF THE APP 2 IN EMULATOR ON MOVESCOUNT.COM.

The latest example is the fartlek training guided by hearth rate where exercise is divided into three laps. At first the application leads athlete to warm up by jog for at least five minutes and to reaching 60% of HRmax. Second lap is then fartlek run where the application leads athlete by current heart rate to acceleration (instruction "RUN +") up to 75% HRmax then to calm down (instruction "JOG") to below 60% HRmax and subsequently to further repeats of the cycle for 20 min. The last sequence (lap) is then calming (jog) for 5 minutes.

if (SUUNTO_LAP_NUMBER == 0) {RESULT = 0; prefix = "STAR"; postfix = "T";} if (SUUNTO_LAP_NUMBER == 1 && SUUNTO_LAP_DURATION < 300) {RESULT = 300-SUUNTO_LAP_DURATION; prefix = "JOG"; postfix = "s";} if (SUUNTO_LAP_NUMBER == 1 && SUUNTO_HR_AVG[30] < 0.6*SUUNTO_USER_MAX_HR && SUUNTO_LAP_DURATION > 300) {RESULT = 0.6*SUUNTO_USER_MAX_HR; prefix = "RUN+"; postfix = "!";} if (SUUNTO_LAP_NUMBER == 1 && SUUNTO_HR_AVG[30] >= 0.6*SUUNTO_USER_MAX_HR && SUUNTO_LAP_DURATION > 300) {RESULT = 2; prefix = "LAP"; postfix = "Go"; Suunto.alarmBeep();}

if (SUUNTO_LAP_NUMBER == 2 && SUUNTO_LAP_DURATION < 1200) {if (INTERVAL == 0 && SUUNTO_HR < 0.75*SUUNTO_USER_MAX_HR) {RESULT = 0.75*SUUNTO_USER_MAX_HR; prefix = "RUN+"; postfix = "!";} if (INTERVAL == 0 && SUUNTO_HR >= 0.75*SUUNTO_USER_MAX_HR) {RESULT = 0; prefix = "SLOW"; postfix = "!";



Suunto.alarmBeep(); INTERVAL = 1;} if (INTERVAL == 1 && SUUNTO_HR > 0.6*SUUNTO_USER_MAX_HR) {RESULT = 0.6*SUUNTO_USER_MAX_HR; prefix = "SLOW"; postfix = "!";} if (INTERVAL == 1 && SUUNTO_HR <= 0.6*SUUNTO_USER_MAX_HR) {RESULT = 0; prefix = "RUN+"; postfix = "!"; Suunto.alarmBeep(); INTERVAL = 0;}}

if (SUUNTO_LAP_NUMBER == 2 && SUUNTO_LAP_DURATION >= 1200) {RESULT = 3; prefix = "LAP"; postfix = "Go"; Suunto.alarmBeep();}

if (SUUNTO_LAP_NUMBER == 3 && SUUNTO_LAP_DURATION <= 300) {RESULT = (300 - SUUNTO_LAP_DURATION); prefix = "JOG"; postfix = "s";} if (SUUNTO_LAP_NUMBER == 3 && SUUNTO_LAP_DURATION > 300) {RESULT = 0; prefix = "STOP"; postfix = "!"; Suunto.alarmBeep();} if (SUUNTO_LAP_NUMBER > 3){RESULT = 0; prefix = "STOP"; Suunto.alarmBeep();}

FIGURE III: PREVIEW OF THE APP 3 IN EMULATOR ON MOVESCOUNT.COM.

V. CONCLUSION All Suunto Apps presented here have been tested by authors on Suunto Ambit 2 device and

published in the App Zone portal movescount.com under the names:

MINIMUM HR LOCATOR 2 [11] FIND MAXIMAL HR - ENDURANCE RUNNERS [12] FARTLEK TRAINING [13]

So far no systematic evaluation of the usability of such applications has been made and at this time the authors don´t know about any general inquiry about the benefits of using Suunto Apps between sport users (the first devices/firmware supporting Monkey C apps are planned by Garmin to enter market in year 2015). Based on our own experience and the available records of



www.elvedit.com 388

each exercise created with use of application on the portal movescount.com we can formulate at least some subjective views and experiences.

It is obvious that Suunto Apps have many limitations. First limitation is simplicity and small range of Suunto Apps scripting language. Authors plan to program similar apps also in Monkey C and test it on Garmin devices to compare both competitive solutions. But main obstacle for development is that the solution (by Suunto and also one introduced by Garmin) is highly proprietary. App created in Suunto App environment is applicable only in Suunto highest product line Ambit. For the owners of compatible devices apps are interesting and effectively usable feature.

To create an effective application certain combinations (at least minimal) of programming skills and knowledge of sports training is required. But in good combination of both it can result in very effective extension of functionality of watch. Number of additional features and displayed parameters can be prepared and especially specific apps can be tailored to lead athletes or entire training groups through trainings (assuming the use of compatible devices). This can be particularly advantageous if athletes are not training under the direct guidance of coach or they want to fully concentrate on their own performance and not on the next steps of exercise.

ACKNOWLEDGMENT This work was supported in part by internal grant Utilization of Mobile Devices in Amateur

Sport and by individual academic support of College of Polytechnics Jihlava.

REFERENCES [1] Garmin Blog. Retrieved from:

http://garmin.blogs.com/my_weblog/2014/09/-introducing-connect-iq-the-first-ever-open-platform-for-garmin-wearables.html#.VFXvhfmG_1Y (quoted 26. 9. 2014).

[2] Baum, P., Abadie, F. Market Developments – Remote Patient Monitoring and Treatment, Telecare, Fitness/Wellness and mHealth. (p.35). Luxembourg: Publications Office of the European Union, 2013. ISBN 978-92-79-25708-7.

[3] Parker, S. History of Heart Rate Monitors. Retrieved from: http://www.articlesbase.com/health-articles/history-of-heart-rate-monitors-253755.html (quoted 29.7.2014).

[4] Polar.com. Retrieved from: http://www.polar.com/en/about_polar/who_we_are/innovations (quoted 29.7.2014)

[5] Máček, M., Radvanský J. Fyziologie a klinické aspekty pohybové aktivity: Fyziologie tělesné zátěže. Praha: Galén, 2011. ISBN 9788072626953.

[6] Máček, M., Radvanský, J. Fyziologie a klinické aspekty pohybové aktivity: Mechanismy působení pohybové aktivity, její nedostatek, detrénink. Praha: Galén, 2011. ISBN 9788072626953.

[7] Benson, R., Connolly, D. Heart rate training. Champaign: Human Kinetics, 2011. ISBN 978-0-7360-8655-4



[8] Suunto Apps, Developer manual. Version February 2014. Retrieved from: http://dcjitw11p57ya.cloudfront.net/downloads/SuuntoAppZoneDeveloperManual6206759C5CFCCF15E08641C96E125185.pdf

[9] Ambit Smart Application. Retrieved from: http://www.movescount.com/groups/group5505-Ambit_Smart_Applications (quoted 31.8.2014)

[10] Connect IQ. Retrieved from: http://developer.garmin.com/connect-iq/overview/ (quoted 4.10.2014)

[11] MINIMUM HR LOCATOR 2. Retrieved from: http://www.movescount.com/apps/app10043382-MINIMUM_HR_LOCATOR_2

[12] FIND MAXIMAL HR - ENDURANCE_RUNNERS. Retrieved from: http://www.movescount.com/apps/app10041528-FIND_MAXIMAL_HR_-_ENDURANCE_RUNNERS

[13] FARTLEK TR. Retrieved from: http://www.movescount.com/apps/app10043525-FARTLEK_TR


Jakub Novotny was born in Jihlava, Czech Republic, studied economics and informatics at University of Economics in Prague and has received Ph.D. degree from applied informatics at this university. The topic of his Ph.D. dissertation was Business metadata and their use in the design of data warehouse. He works now at College of Polytechnics Jihlava in Department of Electrical Engineering and Computer Science. The main areas of his academic interests are effectiveness evaluation of investment in ICT, business

information systems and use of ICT in sport.

Martina Winklerová was born in Ostrov nad Ohří, Czech Republic, studied sport and English language didactics at University of Hradec Kralove and kinantropology at Masaryk University Brno. She has received Ph.D. degree from kinatropology on Masaryk University Brno. The topic of her Ph.D. dissertation was The Influence of Different Types of Physical Education Lessons on Current Mood State of Adolescents. She works now at College of Polytechnics Jihlava in Department of Languages and cooperates also with

Department of Sports. The main areas of her academic interests are kinatropolgy and English language.

IJACSIT Volume 03 - ELVEDIT.COM

Documents

Transcript of IJACSIT Volume 03 - ELVEDIT.COM