CRM RFM 客户价值

ClassifyingthesegmentationofcustomervalueviaRFMmodelandRStheoryChing-HsueCheng,You-ShyangChen*DepartmentofInformationManagement,NationalYunlinUniversityofScienceandTechnology,123,Section3,UniversityRoad,Touliu,Yunlin640,TaiwanarticleinfoKeywords:CRM(customerrelationshipmanagement)CustomervalueanalysisK-meansalgorithmRoughsettheoryRFM(recency,frequencyandmonetary)modelabstractDataminingisapowerfulnewtechniquetohelpcompaniesminingthepatternsandtrendsintheircus-tomersdata,thentodriveimprovedcustomerrelationships,anditisoneofwell-knowntoolsgiventocustomerrelationshipmanagement(CRM).However,therearesomedrawbacksfordataminingtool,suchasneuralnetworkshaslongtrainingtimesandgeneticalgorithmisbrutecomputingmethod.Thisstudyproposesanewprocedure,joiningquantitativevalueofRFMattributesandK-meansalgorithmintoroughsettheory(RStheory),toextractmeaningrules,anditcaneffectivelyimprovethesedraw-backs.Threepurposesinvolvedinthisstudyinthefollowing:(1)discretizecontinuousattributestoenhancetheroughsetsalgorithm;(2)clustercustomervalueasoutput(customerloyalty)thatisparti-tionedinto3,5and7classesbasedonsubjectiveview,thenseewhichclassisthebestinaccuracyrate;and(3)findoutthecharacteristicofcustomerinordertostrengthenCRM.ApracticalcollectedC-companydatasetinTaiwan’selectronicindustryisemployedinempiricalcasestudytoillustratetheproposedprocedure.Referringto[Hughes,A.M.(1994).Strategicdatabasemarket-ing.Chicago:ProbusPublishingCompany],thisstudyfirstlyutilizesRFMmodeltoyieldquantitativevalueasinputattributes;next,usesK-meansalgorithmtoclustercustomervalue;finally,employsroughsets(theLEM2algorithm)tomineclassificationrulesthathelpenterprisesdrivinganexcellentCRM.Inanalysisoftheempiricalresults,theproposedprocedureoutperformsthemethodslistedintermsofaccuracyrateregardlessof3,5and7classesonoutput,andgeneratesunderstandabledecisionrules.2008ElsevierLtd.Allrightsreserved.1.IntroductionDuetothecomplicationanddiversificationofbusinessopera-tion,informationofcompanyisessentialandvitalforcesforadvantagecompetitionandgoing-concern.Particularly,thegrow-ingofinformationtechnology(IT)inrapidchangingandcompeti-tiveenvironmenttodaymotivatestheactivityoftransaction,whichincreasinglyfacilitiesthemarketscompetition.Basedonthisrelationship,informationservesascentraltofacetheopportu-nitiesandchallengesofday-to-dayoperationforcompanies.Itisverydifficultforcompaniesthatstrengthenbusiness’scompetitiveadvantageifinformationonlybecomestosupportthefunctionswithincompanywhenfacingtotheheavychallengescomingfromoutsidessurroundings.Thus,howtoenhancethemarketcompet-itivepowerforcompaniesisaninterestingissuebecauseofthemorethecompetitivepower,themoretheprobabilityforgoing-concern.ThekeypointgainingprofitofcompaniesistointegratetheupstreammembersofsupplychainviaaneffectiveITinordertoreducecost,andreinforcethedownstreamcustomerrelation-shipsviaanexcellentCRMinordertogainmoreprofit.CRMbe-comesthefocalpointofcompanyprofitsandmoreandmoreimportantforcompaniesbecausecustomersaremainresourcesofprofits.Therefore,thisstudyinsistsonthatanexcellentCRMwithcustomersforcompaniesisacriticalforgainingmoreprofit.Thefulfillmentofcustomerrequirementsisoneofkeyfactorsforthesuccessofbusinessoperation.CRMistoachievetheneedsofcustomersandtoenhancethestrengthwithcustomersforcom-pany(Thompson&Sims,2002).However,theeffectiveandeffi-cientutilizationofITtosupporttheCRMprocessisshortpathforsuccessfulCRM.Althoughunderstandingthesituationsofcus-tomersissomewhatdifferent,thecompaniesthatallprovideprod-uctsandservicesforcustomerstosatisfytheirdemandsaresimilartominevaluableinformationofcustomers,torealizethecustomervaluemaximization,toincreasecustomerloyaltyandfinallytoob-tainplentyprofitsforthemselves(Joo&Sohn,2008).Therefore,alargenumberofcompaniesapplythedifferenttoolssuchascom-putersoftwarepackage,statisticaltechniques,toenhanceamoreefficientCRM,inordertoletcompaniesunderstandingmoreabouttheircustomers.Nowadays,byutilizingdataminingtoolsforassistingCRM,sometechniques,whichincludedecisiontrees(DT),artificialneuralnet-works(ANN),geneticalgorithms(GA),associationrules(AR),etc.,areusuallyusedinsomefieldssuchasengineering,science,finance,business,tosolverelatedproblemswithcustomers(Witten&Frank,2005).Adecisiontreeisaflow-chart-liketreestructure,whereeach0957-4174/$-seefrontmatter2008ElsevierLtd.Allrightsreserved.doi:10.1016/j.eswa.2008.04.003*Correspondingauthor.E-mailaddress:[e-mail](Y.-S.Chen).ExpertSystemswithApplications36(2009)4176–4184ContentslistsavailableatScienceDirectExpertSystemswithApplicationsjournalhomepage:www.elsevier.com/locate/eswainternalnodedenotesatestonanattribute,eachbranchrepresentsanoutcomeofthetest,andleafnodesrepresentclassorclassdistri-butions(Han&Kamber,2001).Anartificialneuralnetworkisalargenumberofhighlyinterconnectedprocessingelements(neurons)thatusesamathematicalmodel,computationalmodelornon-linearstatisticaldatamodelingtoolsforinformationprocessingtocaptureandrepresentcomplexinput/outputrelationships.Geneticalgo-rithms,whichwereformallyintroducedintheUnitedStatesinthe