CUSTOMER SEGMENTATION THROUGH FUZZY C-MEANS AND FUZZY RFM METHOD.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Jour nal of The oretical a nd Applied Information Tec hnology

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

JOURNAL OF THEORETICAL AND APPLIED

INFORMATION TECHNOLOGY

EDITORIAL COMMITTEE

NIAZ AHMAD

(Chief Editor)

Professor, FCE, MOE, H-9 Islamabad

PAKISTAN

SHAHBAZ GHAYYUR

(Co- Chief Editor)

Assistant Professor, DCS, FBAS, International Islamic University Islamabad,

PAKISTAN

SAEED ULLAH

(Associate Editor)

Assistant Professor, DCS, Federal Urdu University of Arts, Science & Technology Islamabad,

PAKSITAN

MADIHA AZEEM

(Associate Editor)

Journal of Theoretical and Applied Information Technology, Islamabad.

PAKISTAN

SALEHA SAMAR

(Managing Editor)

PAKISTAN

KAREEM ULLAH

(Managing Editor)

PAKISTAN

SHAHZAD A. KHAN

Lecturer IMCB, FDE Islamabad, PAKISTAN

(Managing Editor/Linguists & In-charge Publishing)

PAKISTAN

August 2015. Vol. 78 No.3 .

i

(10)

Jour nal of The oretical a nd Applied Information Tec hnology

JOURNAL OF THEORETICAL AND APPLIED

REGIONAL ADVISORY PANEL

Dr. SIKANDAR HAYAT KHIYAL

Professor &Chairman DCS& DSE, Fatima Jinnah Women University, Rawalpindi, PAKISTAN

Dr. MUHAMMAD SHER

Professor &Chairman DCS, FBAS, International Islamic University Islamabad, PAKISTAN

Dr. ABDUL AZIZ

Professor of Computer Science, University of Central Punjab, PAKISTAN

Dr. M. UMER KHAN

Asst. Professor Department of Mechatronics, Air University Islamabad, PAKISTAN

Dr. KHALID HUSSAIN USMANI

Asst. Professor Department of Computer Science, Arid Agriculture University,

Rawalpindi, PAKISTAN

August 2015. Vol. 78 No.3

.

ii

(11)

Jour nal of The oretical a nd Applied Information Tec hnology

JOURNAL OF THEORETICAL AND APPLIED

ED ITORIAL ADVISORY BOARD

.

iii Dr. CHRISTEL BAIER

Technical University Dresden, GERMANY

Dr KHAIRUDDIN BIN OMAR UniversitiKebangsaanMalysia, 43600 Bangi

Selangor Darul-Ehsan, MALYSIA

Dr. YUSUF PISAN

University of Technology, Sydney, AUSTRALIA

Dr. S. KARTHIKEYAN Department of Electronics and Computer Engineering, Caledonian College of Engineering,

OMAN (University College with Glascow University, Scotland, UK) DR. YUXIN MAO

School Of Computer & Information Engineering Zhejiang Gongshang University, CHINA

Dr. ZARINA SHUKUR

FakultiTeknologidanSainsMaklumat, University Kebangsaan MALYSIA

Dr. NOR AZAN MAT ZIN

Faculty of Information Science & Technology, National University of MALYSIA

Dr. R.PONALAGUSAMY

National Institute of Technology, Tiruchirappalli, Tamil Nadu, INDIA

Dr. MOHAMMAD TENGKU SEMBOK Universiti Kebangsaan MALYSIA

Dr. PRABHAT K. MAHANTI University of New Brunswick, Saint John, New

Brunswick, CANADA

Dr. NITIN UPADHYAY

Birla Institute of Technology and Science (BITS), Pilani-Goa Campus, INDIA

Dr. S.S.RIAZ AHAMED

Mohamed Sathak Engineering College, Kilakarai, &Sathak Institute of Technology, Ramanathapuram , Tamilnadu, INDIA

Dr. A. SERMET ANAGÜN

Eskisehir Osmangazi University, Industrial Engineering Department, Bademlik Campus,

26030 Eskisehir, TURKEY.

Dr. YACINE LAFIFI

Department of Computer Science, University of Guelma, BP 401, Guelma 24000, ALGERIA.

Dr. CHRISTOS GRECOS

School Of Computing, Engineering And Physical Sciences University Of Central Lancashire.

UNITED KINGDOM

Dr. JAYANTHI RANJAN Institute of Management Technology Raj Nagar, Ghaziabad, Uttar Pradesh, INDIA

Dr. ADEL M. ALIMI

National Engineering School of Sfax (ENIS), University of SFAX, TUNISIA

Dr. RAKESH DUBE

Professor & Head, RKG Institute of Technology, Ghaziabad, UP, INDIA

Dr. ADEL MERABET

Department of Electrical & Computer Engineering, Dalhousie University, Halifax,

CANADA

Dr. HEMRAJ SAINI

CE&IT Department, Higher Institute of Electronics, BaniWalid. LIBYA

(12)

Jour nal of The oretical a nd Applied Information Tec hnology

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 August 2015. Vol. 78 No.3

.

iv Dr. SEIFEDINE KADRY

Lebanese International University, LEBONON

Dr. AIJUAN DONG Department of Computer Science Hood College Frederick, MD 21701. USA

Dr. ZURIATI AHMAD ZUKARNAIN University Putra Malaysia,

MALAYSIA

Dr. HEMRAJ SAINI

Higher Institute of Electronic, BaniWalid LIBYA

Dr. CHELLALI BENACHAIBA University of Bechar, ALGERIA

Dr. MOHD NAZRI ISMAIL

University of Kuala Lumpur (UniKL) MALYSIA

Dr. VITUS SAI WA LAM The University of Hong Kong, CHINA

Dr. WITCHA CHIMPHLEE SuanDusitRajabhat University, Bangkok,

THAILAND

Dr. SIDDHIVINAYAK KULKARNI University of Ballarat, Ballarat,

AUSTRALIA

Dr. S. KARTHIKEYAN Caledonian College of Engineering,

OMAN

Dr. DRAGAN R. MILIVOJEVIĆ

Mining and Metallurgy Institute BorZelenibulevar 35, 19210 Bor, SERBIA

Dr. E. SREENIVASA REDDY Principal - VasireddyVenkatadri Institute of

Technology, Guntur, A.P., INDIA

Dr OUSMANE THIARE Gaston Berger University, Department of Computer Science, UFR S.A.T, BP 234 Saint-

Louis SENEGAL

Dr. SANTOSH DHONDOPANT KHAMITKAR RamanandTeerthMarathwada University, Nanded.

Maharashtra 431605, INDIA

Dr. M. IQBAL SARIPAN

(MIEEE, MInstP, Member IAENG, GradBEM) Dept. of Computer and Communication Systems Engineering, Faculty of Engineering, Universiti

Putra MALAYSIA

Dr. E. SREENIVASA REDDY Principal - VasireddyVenkatadri Institute of

Technology, Guntur, A.P., INDIA

Dr. T.C.MANJUNATH, Professor & Head of the Dept., Electronicis& Communication Engg. Dept,

New Horizon College of Engg., Bangalore-560087, Karnataka, INDIA.

Dr. SIDDHIVINAYAK KULKARNI

Graduate School of Information Technology and Mathematics University of Ballart AUSTRALIA

Dr. SIKANDAR HAYAT KHIYAL Professor & Chairman DCS& DSE, Fatima

Jinnah Women University, Rawalpindi,

PAKISTAN

Dr. MUHAMMAD SHER Professor & Chairman DCS, FBAS, International Islamic University Islamabad,

PAKISTAN

Dr. ABDUL AZIZ

Professor of Computer Science, University of Central Punjab, PAKISTAN

Dr. M. UMER KHAN

(13)

Jour nal of The oretical a nd Applied Information Tec hnology

Elite Panel Members Have A Decision Weight Equivalent of Two Referees (Internal OR External). The Expertise Of Editorial Board Members Are Also Called In For Settling Refereed Conflict About

Acceptance/Rejection And Their Opinion Is Considered As Final.

.

v Dr. RIKTESH SRIVASTAVA

Assistant Professor, Information Systems Skyline University College P O Box 1797, Sharjah, UAE

Dr. BONNY BANERJEE

PhD in Computer Science and Engineering, The Ohio State University, Columbus, OH, USA

Senior Scientist Audigence, FL, USA

PROFESSOR NICKOLAS S. SAPIDIS DME, University of Western Macedonia

Kozani GR-50100, GREECE.

Dr. NAZRI BIN MOHD NAWI Software Engineering Department, Faculty of

Science Computer Information Technology,

Universiti Tun Hussein Onn

MALAYSIA

Dr. JOHN BABALOLA OLADOSU Ladoke Akintola University of Technology,

Ogbomoso, NIGERIA

Dr. ABDELLAH IDRISSI

Department of Computer Science, Faculty of Science, Mohammed V University - Agdal,

Rabat, MOROCCO

Dr. AMIT CHAUDHRY

University Institute of Engineering and Technology, Panjab University, Sector-25,

Chandigarh, INDIA

Dr. ASHRAF IMAM

Aligarh Muslim University, Aligarh-INDIA

Dr. MOHAMMED ALI HUSSAIN Dept. of Computer Science & Engineering, Sri Sai Madhavi Institute of Science & Technology,

Mallampudi, Rajahmundry, A.P, INDIA

Dr. KHALID HUSSAIN USMANI Asst. Professor Department of Computer Science,

Arid Agriculture University, Rawalpindi, PAKISTAN

Dr. GUFRAN AHAMD ANSARI Qassim University, College of Computer Science,

Ministry of Higher Education, Qassim University, KINGDOM OF SAUDI

ARABIA

Dr. Defa Hu

School of Information, Hunan University of Commerce

Changsha 410205, Hunan, P. R. of China

(14)

Jour nal of The oretical a nd Applied Information Tec hnology

PREFACE

Journal of Theoretical and Applied Information Technology (JATIT) published since 2005 (E-ISSN 1817- 3195 / ISSN 1992-8645) is an International refereed research publishing journal with a focused aim of promoting and publishing original high quality research dealing with theoretical and scientific aspects in all disciplines of Information Technology. JATIT is an international scientific research journal focusing on issues in information technology research. A large number of manuscript inflows, reflects its popularity and the trust of world's research community. JATIT is indexed with various organizations and is now published on monthly basis.

All technical or research papers and research results submitted to JATIT should be original in nature, never previously published in any journal or undergoing such process across the globe. All the submissions will be peer-reviewed by the panel of experts associated with JATIT. Submitted papers should meet the internationally accepted criteria and manuscripts should follow the style of the journal for the purpose of both reviewing and editing. All of its articles also appear online as per policy of JATIT

Journal of Theoretical and Applied Information Technology receives papers in continuous flow and we will consider articles from a wide range of Information Technology disciplines encompassing the most basic research to the most innovative technologies. Please submit your papers electronically to our submission system at http://jatit.org/submit_paper.php in an MSWord, Pdf or compatible format so that they may be evaluated for publication in the upcoming issue. This journal uses a blinded review process; please remember to include all your personal identifiable information in the manuscript before submitting it for review, we will edit the necessary information at our side. Submissions to JATIT should be full research / review papers (properly indicated below main title).

It is the sole responsibility of the submitting authors to make sure that the submitted manuscript is not in process of publication anywhere in any conference/journal across the globe, nor part or whole of it is copied from any source.

The review process may take anywhere from five days to two months depending on the response time to referees. Authors will be informed about the updated status via e-mail as soon as we receive the evaluation results. After submission of publication dues for accepted manuscripts a publication slot will be allocated to your manuscript for its publication in upcoming monthly issues of JATIT.

******************

.

(15)

Jour nal of The oretical a nd Applied Information Tec hnology

ABSTRACTING & INDEXING

Journal of Theoretical and Applied Information Technology Islamabad Pakistan is focused, double blind peer reviewed journal that is now being published monthly and is published by Asian Research Publishing Network and is Indexed / Abstracted by the following International Agencies and institutions. JATIT has been regularly published since 2005 and now has a well reputed international standing and invites contributions from researchers, scientists, and practitioners from all over the world.

*- Ulrich's Periodicals Directory

*- DataBase systems and Logic Programming (DBLP) *- EBSCO Publishing USA

*- Directory of Open Access Journals (DOAJ) *- Google & Google Scholar Journals

*- The Index of Information Systems Journals *- Information Technology Resources Collection *- ZDNet Australia

*- NLM Catalog

*- Computing Research and Education Association of Australasia *- CiteSeer

*- Elsevier *- SCOPUS

*- Engineering Village *- TOC Premier

******************

Feel free to suggest JATIT to any Indexing & Abstracting Services which are appropriate to its scope

TM

*- Computer Science Journals

*- Computers and Applied Sciences Complete *- N|W Switzerland

*- Microsoft Academic Search *- Cabell Publishing

*-OpenJgate *- INSPEC

*- IAOR Palgrave Macmillan

.

vii

(16)

Journal of Theoretical and Applied Information Technology

31stAugust 2015. Vol.78. No.3

ISSN:1992-8645 www.jatit.org E-ISSN:1817-3195

380

CUSTOMER SEGMENTATION THROUGH

FUZZY C-MEANS AND FUZZY RFM METHOD

NI PUTU PUTRI YULIARI1,I KETUT GEDE DARMA PUTRA2, NI KADEK DWI RUSJAYANTI3

Department Of Information Technology, Engineering Faculty in Udayana University Bukit Jimbaran, Bali, Indonesia, Telp. +62361703315

E-mail: putriyuliari@gmail.com1,Ikgdarmaputra@gmail.com2,dwi.rusjayanti@gmail.com3

ABSTRACT

This research aims for finding the potential customer use data transaction. This causes, the company is difficult to arrange customers who have high and low loyalty and this research have a application for customer segmentation to help analyzing transaction data with Fuzzy C-Means for clustering and Fuzzy RFM for identify the customers. Softwares used to conduct this experiment are Microsoft SQL Server to saving the database and Matlab as the tools. The results of this segmentation for four experiments are two classes. Its has superstar I and Occasional H for each number cluster and then for the best number of cluster for this experiments are two clusters according MPC method.

Keywords: Customer segmentation, Clustering, Fuzzy c-means, Fuzzy RFM, MPC (Modified Partition

Coefficient)

1. INTRODUCTION

The company is difficult to arrange customers who have high and low loyalty. This is caused of data transaction growth fast and limited ability to segmenting with manual computation. Amount of data from furniture company has important information to segmenting, the process of finding in a set of data called data mining[1]. Data mining is used to give services to customers based on views or insights of customers with CRM strategy. Every relation with customers can make a benefit to the company. Profitable relationship is done by analyzing data transaction of customers. Because of that marketing is important to dividing customers[2]. Volume data is continues to grow and can not be analysis with manual[1]. The application of data mining can help in analyzing customers to determine level of loyalty customer.

Effective segmentation leads to competitive advantage, recognition and exploitation of new market opportunities, selection of the appropriate target market, enhanced differentation and positioning, and increased profitability. Despite the appealing strategic and tactical benefit of market segmentation, cluster analysis remain the most favoured method.[6] The basic of idea of cluster analisys is to divide a heterogeneous customers market into homogeneous sub-grups[9]. But, some information is inevitably lost when object are grouped. Information loss is

not problematic but it can result in the wrong conclusions[8]. Hence, there is no succesful segmentation without an appropriate clustering algorithm[7].

Therefore, this research have a application for customer segmentation to help analyzing transaction data in a furniture company, the application is developing method of Fuzzy C-Means and Fuzzy RFM. Software used to conduct this experiment is Microsoft SQL Server for saving database and Matlab as the tools. This application used fuzzy clustering algorithm with Fuzzy C-Means method, the algorithm have been selected because this method can make data grouped by the cluster. Fuzzy RFM (Recency, frequency, monetary) method used to choose customer with high or low loyalty from the result data of Fuzzy C-Means method. Fuzzy RFM can determine customer to the class with level loyalty their have.

2. CUSTOMER SEGMENTATION

(17)

Journal of Theoretical and Applied Information Technology

381 good cluster and characteristic distinctly.

Figure 1 explaining about system process for customer segmentation, the input of this system is database from the company. Database from the company was choosed use data preparation process, data will divided to third group with Fuzzy RFM parameter. Data with paramater will be clustering with Fuzzy C-Means method and then apply MPC method for validity cluster. Output from this application is class category of the customer.

Figure 1: System Process

3. CLUSTERING

The RFM method used for determine variable of measuring purchase products by customers. Variable can determine as recency, frequency and monetary[3].

a. Recency is range time (day, month, year) from end transaction until this time by customers purchase.

b. Frequency is transaction total or transaction average in once period.

c. Monetary is average cost total of customers in once period.

Segmentation cluster in the retail company divided by six characteristic with RFM values of customers[2].

Table 1:Customer Characteristic with RFM Values.

Customer Class Description

Superstar a. Customers with high loyalty

b. High values

c. High frequency values d. Highest transaction

Golden Customer a. Second high values

b. High frequency values c. Standard transaction values

Typical Customer a. Have standard value and

transaction values

Occational Customer a. Second of the last frequency

values after dormant customer b. Lowest recency

c. Highest transaction

Everyday Shopper a. Have raising transaction

b. Low transaction

c. Have value with middle until low scale

Dormant Customer a. Lowest frequency and value

b. Lowest recency

Attribute distribution base on RFM will show in Table 2.

Table 2: Domain Value RFM

Atribute Linguistic variable

Domain Value

Recency Long Time Ago

Rather Longer Recently

0≤ r < Max_r1 day

Max_r1 day< r < Max_r2 day

Max_r2 day < r

Frequency Seldom

Rather Frequent Often

Very Often

0≤ f < Max_f1 transaction

Max_f1 transaction < f < Max_f2 transaction Max_f2 transaction <f< Max_f3 transaction Max_f3 transaction < f

Monetary Very Low

Low Rather Low Rather High High Very High

0≤ m < Max_m1 Rupiah

IDR Max_m1 < m < IDR Max_m2

IDR Max_m5 < m

Fuzzy RFM used trapezoid graph for dispart the domain value. The graph of domain value from fuzzy RFM will show in figure 2.

Figure 2a : Fuzzy RFM Recency

(18)

Journal of Theoretical and Applied Information Technology

382

Figure 2c : FuzzyRFM Recency

Customer segmentation process will do with computing membership degree of the centroid from every cluster with all class from fuzzy model. Its computing used equation from zumstein[11]

(1) Explanation:

µA = Degree of membership for every class

µi = Degree of membership for every linguistic

variable in Fuzzy RFM A = Class in RFM Model i = Linguistic Variable x = Centroid

γ = Gamma, using value 0,5

Table 2 explain about limit of class for customer segmentation as superstar, golden, typical, occasional, everyday and dormant. The FCM approach has been applied to data clustering. Then for validation test of data using MPC method, its has for make sure the best number of clusters. The algorithm of MPC method is[10]:

(2) The C value is the centroid and then MPC(c) is index value of MPC when cluster have c

value.

4. RESULT AND DISSCUSION

4.1 Arsithecture Data Analysis

Arsithecture data for customer segmentation divided to 3 part which are data selection, preprocessing and transformation. Data selection is using transaction data of furniture company. Preprocessing step base on RFM method used customerID, order date and unit price from database. And the final step is transformation data to RFM.

Figure 3 : Step of Architechture Data Selection

Matlab R2014b application has applied for implement FCM method. The ODBC is used for make relation between Matlab and SQL server. If Matlab and SQL server has a relation then the selection data can do based RFM method.

Figure 4 : RFM Data in 3D Graphic

Figure 4 explaining about data after execute to RFM and the graphic have information about dissemination data of the company. User can input total cluster, weight and maximum iteration.

4.2 Experiment

(19)

Journal of Theoretical and Applied Information Technology

383 (a)

(b)

(c)

(d)

Figure 5: Dissemination Data Graph (a) 2 Clusters, (b) 3 Clusters, (d) 4 Clusters, (e) 5 Clusters

The result of the clustering process will used to find class of customer with Fuzzy RFM method, before finding the class use Fuzzy RFM. We should to implement MPC method for validition cluster to make sure the result of the clustering is right. We have the best number of cluster according MPC method is two clusters.

Figure 7: Cluster Comparison

The best value for number of cluster (close to one in range 0-1) with Fuzzy RFM method, it means quality of the cluster is more good for clustering data. The comparison of number of cluster with MPC Method showed in figure 7 which are 2, 3, 4 and 5 clusters and the best result is two clusters with value of validation test is closed to one. This comparison aims is for knowing number of cluster with good result for dissemination data.

An error in the usage of cluster number will cause an error in class identification and will affect the marketing strategy of a product in the company which will cause loss in company profit.

(20)

Journal of Theoretical and Applied Information Technology

384

Table 3: The Classification Result Using FCM with 2 Number Cluster

CustomerID Cluster 1 Cluster 2 Classification Result

925 0.9999 0.0001 1.0000

23 0.1029 0.8971 2.0000

687 0.9303 0.0697 1.0000

902 0.9994 0.0006 1.0000

710 0.9994 0.0006 1.0000

879 0.9971 0.0029 1.0000

733 0.9999 0.0001 1.0000

856 0.9996 0.0004 1.0000

779 0.9995 0.0005 1.0000

802 0.9998 0.0002 1.0000

In this case, use the value of cluster center for compared with Fuzzy RFM Method. The cluster center of two clusters shown in Table 4.

Table 4: Value of the Cluster Center

Class Code R F M Score

K26 1 1 0.9641 0.9819

K30 1 1 1 1

In the table 4, class code K30 is code for occasional H class and K26 for superstar I. Superstar customer is the customer with high loyalty, and the customer with Occasional class has low frequency. So, in this research we have two customer characters that is high loyalty (Superstar) and low frequency but highest transaction (Occasional).

Figure 6: The Result of Customer Identified by Fuzzy RFM Method

This research has two result, that is class of the customers and the best number of clusters. the application result can knowing potential customer from class their have. Customers with included to cluster 1 will have Superstar class, and cluster 2 is Occasional class. From the result, the company can make a decision for subjected their customers. For the next experiment according Fuzzy C-Means and Fuzzy RFM method should use two clusters.

5. CONCLUSIONS

The customer segmentation application is developing method of Fuzzy C-Means and Fuzzy RFM with data transaction. Its has can build cluster with superstar customers class from comparing membership degree of the centroid with the class of Fuzzy RFM Method. The result of four experiment with same weight and maximum iteration have two class dominant which are superstar I and Occasional H. The cluster validity test with MPC (Modified Partition Coefficient) have the best number of cluster for FCM is two clusters.

6. FURTHER RESEARCH DIRECTION

The experimental results supported the usefulness of the proposed methodology. In the future, other clustering methods can be fuzzified in similar ways for the same purpose and not only focus on customer segmentation but also determine of the marketing target.

REFRENCES:

[1] Betalya. Konsep Data Mining dan Knowledge

Discovery in Database. Universitas Gunadarma. 2009

[2] Tsiptsis Antonios Chorianopoulos. Data Mining Techniques in CRM: Inside Customer Segmentation. NewCaledonia: Antony Rowe

Ltd, Chippenham, Wiltshire. 2009

[3] Chen. Understanding Customer Relationship

Management (Crm) People, Process And Technology. Vol. 9 No. 5, 2003

[4] Chapman, P., Clinton, J., Kerber, R., Khabaza, T., Reinartz, T., Shearer, C., Wirth, R. 2000.

CRISP-DM 1.0 : Step-by-Step Data Mining Guide. 2000

[5] Ching-hsue, Cheng. Classifying the segmentation of customer value via RFM model and RS theory. Expert Systems with

Applications 36 (2009) 4176–4184.

[6] Dolnicar, S., Crouch, G. I., Devinney, T., Huybers, T., Louviere, J., and Oppewal, H.

Tourism and discretionary income allocation heterogeneity among households. Tourism Management, 2009, 29(1):44–52.

[7] Dolnicar, S. Using cluster analysis for market

(21)

Journal of Theoretical and Applied Information Technology

385 Australasian Journal of Market Research, 2003, 11(2):5–12.

[8] Franke, N., Reisinger, H., and Hoppe, D. Remaining within–cluster heterogeneity: a

meta–analysis of the “dark–side” of clustering methods. Journal of Marketing Management,

2009, 25(3/4):273–293.

[9] Punj, G. and Stewart, D. W. Cluster analysis in

marketing research: review and suggestions for application. Journal of Marketing Research, 1983,20(2):138–148.

[10] Nugraheni, Yohana. Data mining dengan

metode fuzzy untuk customer relationship manangement pada perusahaan retail.

Denpasar : Universitas Udayana. 2011