ISSN: 1992-8645 CUSTOMER SEGMENTATION …

1 Journal of Theoretical and Applied Information Technology31stAugust2015. 2005-2015 JATIT & LLS. All rights :1817-3195380 CUSTOMER SEGMENTATIONTHROUGHFUZZY C-MEANS AND FUZZY RFMMETHODNI PUTU PUTRI YULIARI1,I KETUT GEDE DARMA PUTRA2, NI KADEK DWI RUSJAYANTI3 Department Of Information Technology, Engineering Faculty in Udayana UniversityBukit Jimbaran, Bali, Indonesia,Telp. research aims for finding thepotential CUSTOMER usedata transaction. This causes, thecompany isdifficult to arrange customers who have high and low loyaltyandthis research have a application forcustomer SEGMENTATION to help analyzing transaction datawithFuzzy C-Meansfor clusteringand FuzzyRFMfor used to conduct this experiment are Microsoft SQL Server tosaving the database and Matlab as the tools.

2 The results of this segmentationfor four experiments are twoclasses. Its has superstarIand Occasional H for each number cluster and then for the best number of clusterfor this experiments are two clusters according MPC : CUSTOMER SEGMENTATION , Clustering,Fuzzy c-means, Fuzzy RFM, MPC(Modified PartitionCoefficient) company is difficult to arrangecustomers who have high and low loyalty. This iscaused of data transaction growth fast and limitedability to segmenting with manual of data from furniture company hasimportant information to segmenting, the processof finding in a set of data called data mining[1].Data mining is used to give services to customersbased on views or insights of customers with relation with customers can make abenefit to thecompany.

3 Profitable relationship isdone by analyzing data transaction of of that marketing is important to dividingcustomers[2].Volume data is continues to grow andcan not be analysis with manual[1]. The applicationof data mining can help in analyzing customers todetermine level of loyalty SEGMENTATION leads tocompetitive advantage, recognition andexploitation of new market opportunities, selectionof the appropriate target market, enhanceddifferentation andpositioning, and increasedprofitability. Despite the appealing strategic andtactical benefit of market SEGMENTATION , clusteranalysis remain the most favoured method.[6] Thebasic of idea of cluster analisys is to divide aheterogeneous customers market into homogeneoussub-grups[9].

4 But, some information is inevitablylost when object are grouped. Information loss isnot problematic but it can result in the wrongconclusions[8]. Hence, there is no succesfulsegmentation without an appropriate clusteringalgorithm[7].Therefore,thisres earch have a applicationfor CUSTOMER SEGMENTATION to help analyzingtransaction data ina furniture company, theapplication isdeveloping method of Fuzzy C-Means and Fuzzy used to conductthis experimentis Microsoft SQL Serverforsavingdatabase and Matlab as the tools. This applicationused fuzzy clustering algorithm with Fuzzy C-Means method, the algorithm have been selectedbecause this method can make data grouped by thecluster.

5 Fuzzy RFM (Recency,frequency,monetary) method used to choose CUSTOMER withhigh or low loyalty from the result data of Fuzzy C-Means method. Fuzzy RFM can determinecustomer to the class with level loyalty their SEGMENTATIONS egmentation is process for dividedcustomers to the some cluster with category of theloyalty CUSTOMER for build the market of segmentationaremade bybussiness clustering algorithm can beanalize characteristic of data, cluster identificationand result of monitoring data modelofoperator data miningarebuild for searching theJournal of Theoretical and Applied Information Technology31stAugust2015. 2005-2015 JATIT & LLS. All rights :1817-3195381good cluster and characteristic 1 explaining about system process forcustomer SEGMENTATION , the input of this system isdatabase from thecompany.

6 Databasefrom thecompany waschoosedusedata preparation process,data will divided to third group with Fuzzy RFMparameter. Data with paramater will be clusteringwith Fuzzy C-Means method andthen apply MPCmethod for validity thisapplication is class category of the 1: System method used for determinevariable of measuring purchase products bycustomers. Variable can determine as recency,frequency and monetary[3]. isrange time (day, month, year) fromend transaction until this time by is transaction total or transactionaverage in once is average cost total of customers inonce cluster in the retail company dividedby six characteristic with RFM values ofcustomers[2].

7 Table 1: CUSTOMER Characteristic with RFM with high frequency transactionGolden high frequency transaction valuesTypical standard value andtransaction valuesOccational of the last frequencyvalues after dormant transactionEveryday raising value with middle untillow scaleDormant frequency and recencyAttribute distribution base on RFMwillshow inTable :Domain Value RFMA tributeLinguisticvariableDomain ValueRecencyLong Time AgoRather LongerRecently0 r < Max_r1 dayMax_r1 day< r < Max_r2dayMax_r2 day < rFrequencySeldomRather FrequentOftenVery Often0 f < Max_f1 transactionMax_f1 transaction < f <Max_f2 transactionMax_f2 transaction <f<Max_f3 transactionMax_f3 transaction < fMonetaryVery LowLowRather LowRather HighHighVery High0 m < Max_m1 RupiahIDRMax_m1 < m <IDRMax_m2 IDRMax_m2 < m <IDRMax_m3 IDRMax_m3 < m <IDRMax_m4 IDRMax_m4 < m <IDRMax_m5 IDRMax_m5 < mFuzzy RFM used trapezoid graph for dispartthe domain value.

8 The graph of domain value fromfuzzy RFM will show in figure 2a : Fuzzy RFM RecencyFigure 2b: FuzzyRFM FrequencyJournal of Theoretical and Applied Information Technology31stAugust2015. 2005-2015 JATIT & LLS. All rights :1817-3195382 Figure 2c: FuzzyRFMR ecencyCustomer SEGMENTATION process will dowith computing membership degree of the centroidfrom every cluster with all class from fuzzy computing used equation from zumstein[11](1)Explanation: A= Degree of membership for every class i= Degree of membership for every linguisticvariable in Fuzzy RFMA= Class in RFM Modeli= Linguistic Variablex= Centroid =Gamma,using value 0,5 Table 2 explain about limit of class forcustomer SEGMENTATION as superstar, golden,typical, occasional, everyday and approach has been applied to data for validation test of data using MPC method,its has for make sure the best number of algorithm of MPC method is[10].

9 (2)The C valueis the centroid and thenMPC(c)is index value of MPC when cluster have AND Data AnalysisArsithecture data for customersegmentation divided to 3 part which are dataselection, preprocessing and transaction data of stepbase on RFM methodused customerID, order date and unit price fromdatabase. And the final step is transformation datato 3:Step of Architechture Data SelectionMatlab R2014bapplicationhas applied forimplement FCM method. The ODBC is used formake relation between Matlab and SQL server. IfMatlab and SQL server has a relation then theselection data can do based RFM 4:RFM Data in 3D GraphicFigure4explaining about data after executeto RFM and the graphichave informationaboutdissemination data of the company.

10 User can inputtotal cluster, weight and this section, the experiment is taken todemonstrate with two until five clusters using sameweight and maximal iteration, and then weimplement FCM toclassifythe customers. Thefigures 5 are about dissemination of data base ontotal of Theoretical and Applied Information Technology31stAugust2015. 2005-2015 JATIT & LLS. All rights :1817-3195383(a)(b)(c)(d)Figure 5:Dissemination Data Graph (a) 2 Clusters, (b)3 Clusters, (d) 4 Clusters, (e) 5 ClustersThe result of the clustering process willused to find class of CUSTOMER withFuzzy RFMmethod,before finding the class use Fuzzy should toimplement MPC method forvalidition cluster to make sure the result of theclustering is right.