客戶關(guān)系管理中的數(shù)據(jù)挖掘技術(shù)外文翻譯

客戶關(guān)系管理中的數(shù)據(jù)挖掘技術(shù)外文翻譯畢業(yè)論文（設(shè)計(jì)）外文翻譯外文原文DataminingtechniquesforcustomerrelationshipmanagementChrisRygielski,Jyun-ChengWang,DavidC.YenAbstractAdvancementsintechnologyhavemaderelationshipmarketingarealityinrecentyears.Technologiessuchasdatawarehousing,datamining,andcampaignmanagementsoftwarehavemadecustomerrelationshipmanagementanewareawherefirmscangainacompetitiveadvantage.Particularlythroughdatamining?theextractionofhiddenpredictiveinformationfromlargedatabases?organizationscanidentifyvaluablecustomers,predictfuturebehaviors,andenablefirmstomakeproactive,knowledge-drivendecisions.Theautomated,future-orientedanalysesmadepossiblebydataminingmovebeyondtheanalysesofpasteventstypicallyprovidedbyhistory-orientedtoolssuchasdecisionsupportsystems.Dataminingtoolsanswerbusinessquestionsthatinthepastweretootime-consumingtopursue.Yet,itistheanswerstothesequestionsmakecustomerrelationshipmanagementpossible.Varioustechniquesexistamongdataminingsoftware,eachwiththeirownadvantagesandchallengesfordifferenttypesofapplications.Aparticulardichotomyexistsbetweenneuralnetworksandchi-squareautomatedinteractiondetectionCHAID.Whiledifferingapproachesaboundintherealmofdatamining,theuseofsometypeofdataminingisnecessarytoaccomplishthegoalsof''st(cdajtomerrelationshipmanagementphilosophy.22ElsevierScienceLtd.Allrightsreserved.Keywords:CustomerrelationshipmanagementCRM;Relationshipmarketing;Datamining;Neuralnetworks;Chi-squareautomatedinteractiondetectionCHAID;Privacyrights1.IntroductionAnewbusinesscultureisdevelopingtoday.Withinit,theeconomicsofcustomerrelationshipsarechanginginfundamentalways,andcompaniesarefacingtheneedtoimplementnewsolutionsandstrategiesthataddressthesechanges.Theconceptsofmassproductionandmassmarketing,firstcreatedduringtheIndustrialRevolution,arebeingsupplantedbynewideasinwhichcustomerrelationshipsarethecentralbusinessissue.Firmstodayareconcernedwithincreasingcustomervaluethroughanalysisofthecustomerlifecycle.Thetoolsandtechnologiesofdatawarehousing,datamining,andothercustomerrelationshipmanagementCRMtechniquesaffordnewopportunitiesforddebusinessestoactontheconceptsofrelationshipmarketing.Theoldmddfelsiofndde-build-se'llaproduct-orientedviewisbeingreplacedby“sell-build-redes”gnacustomer-orientedview.Thetraditionalprocessofmassmarketingisbeingchallengedbythenewapproachofone-to-onemarketing.Inthetraditionalprocess,themarketinggoalistoreachmorecustomersandexpandthecustomerbase.Butgiventhehighcostofacquiringnewcustomers,itmakesbettersensetoconductbusinesswithcurrentcustomers.Insodoing,themarketingfocusshiftsawayfromthebreadthofcustomerbasetothedepthofeachcustomerneeds.Theperformancemetricchangesfrommarketsharetoso-calledwalletshar”.Businessesdonotjustdealwithcustomersinordertomaketransactions;theyturntheopportunitytosellproductsintoaserviceexperienceandendeavortoestablishalong-termrelationshipwitheachcustomer.TheadventoftheInternethasundoubtedlycontributedtotheshiftofmarketingfocus.Ason-lineinformationbecomesmoreaccessibleandabundant,consumersbecomemoreinformedandsophisticated.Theyareawareofallthatisbeingoffered,andtheydemandthebest.Tocopewiththiscondition,businesseshavetodistinguishtheirproductsorservicesinawaythatavoidstheundesiredresultofbecomingmerecommodities.Oneeffectivewaytodistinguishthemselvesiswithsystemsthatcaninteractpreciselyandconsistentlywithcustomers.Collectingcustomerdemographicsandbehaviordatamakesprecisiontargetingpossible.Thiskindoftargetingalsohelpswhendevisinganeffectivepromotionplantomeettoughcompetitionoridentifyingprospectivecustomerswhennewproductsappear.Interactingwithcustomersconsistentlymeansbusinessesmuststoretransactionrecordsandresponsesinanonlinesystemthatisavailabletoknowledgeablestaffmemberswhoknowhowtointeractwithit.Theimportanceofestablishingclosecustomerrelationshipsisrecognized,andCRMiscalledfor.ItmayseemthatCRMisapplicableonlyformanagingrelationshipsbetweenbusinessesandconsumers.Acloserexaminationrevealsthatitisevenmorecrucialforbusinesscustomers.Inbusiness-to-businessB2Benvironments,atremendousamountofinformationisexchangedonaregularbasis.Forexample,transactionsaremorenumerous,customcontractsaremorediverse,andpricingschemesaremorecomplicated.CRMhelpssmooththeprocesswhenvariousrepresentativesofsellerandbuyercompaniescommunicateandcollaborate.Customizedcatalogues,personalizedbusinessportals,andtargetedproductofferscansimplifytheprocurementprocessandimproveefficienciesforbothcompanies.E-mailalertsandnewproductinformationtailoredtodifferentrolesinthebuyercompanycanhelpincreasetheeffectivenessofthesalespitch.Trustandauthorityareenhancediftargetedacademicreportsorindustrynewsaredeliveredtotherelevantindividuals.AllofthesecanbeconsideredamongthebenefitsofCRMCapGeminiconductedastudytogaugecompanyawarenessandpreparationofaCRMstrategy[1].Ofthefirmssurveyed,65%wereawareofCRMtechnologyandmethods;28%hadCRMprojectsunderstudyorintheimplementationphase;12%wereintheoperationalphase.In45%ofthecompaniessurveyed,implementationandmonitoringoftheCRMprojecthadbeeninitiatedandcontrolledbytopmanagement.Thus,itisapparentthatthisisanewandemergingconceptthatisseenasakeystrategicinitiative.Thisarticleexaminestheconceptsofcustomerrelationshipmanagementandoneofitscomponents,datamining.ItbeginswithanoverviewoftheconceptsofdataminingandCRM,followedbyadiscussionofevolution,characteristics,techniques,andapplicationsofbothconcepts.Next,itintegratesthetwoconceptsandillustratestherelationship,benefits,andapproachestoimplementation,andthelimitationsofthetechnologies.Throughtwostudies,weofferacloserlookattwodataminingtechniques:Chi-squareAutomaticInteractionDetectionCHAIDandNeuralNetworks.Basedonthosecasestudies,CHAIDandneuralnetworksarecomparedandcontrastedonthebasisoftheirstrengthsandweaknesses.Finalldyr,awweconclusionsbasedonthediscussion.Definition“Datamining”isdefinedasasophisticateddatasearchcapabilitythatusesstatisticalalgorithmstodiscoverpatternsandcorrelationsindata[2].Thetermisananalogytogoldorcoalmining;dataminingfindsandextractsknowledge“datanuggets”buriedincorporatedatawarehouses,orinformationthatvisitorshavedroppedonawebsite,mostofwhichcanleadtoimprovementsintheunderstandinganduseofthedata.Thedataminingapproachiscomplementarytootherdataanalysistechniquessuchasstatistics,on-lineanalyticalprocessingOLAP,spreadsheets,andbasicdataaccess.Insimpleterms,dataminingisanotherwaytofindmeaningindata.Dataminingdiscoverspatternsandrelationshipshiddenindata[3],andisactuallypartofalargerprocesscalled“knowledgediscovery”whichdescribesthestepsthatmustbetakentoensuremeaningfulresults.Dataminingsoftwaredoesnot,however,eliminatetheneedtoknowthebusiness,understandthedata,orbeawareofgeneralstatisticalmethods.Dataminingdoesnotfindpatternsandknowledgethatcanbetrustedautomaticallywithoutverification.Datamininghelpsbusinessanalyststogeneratehypotheses,butitdoesnotvalidatethehypotheses.TheevolutionofdataminingDataminingtechniquesaretheresultofalongresearchandproductdevelopmentprocess.Theoriginofdatamininglieswiththefirststorageofdataoncomputers,continueswithimprovementsindataaccess,untiltodaytechnologyallowsuserstonavigatethroughdatainrealtime.Intheevolutionfrombusinessdatatousefulinformation,eachstepisbuiltonthepreviousones.Table1showstheevolutionarystagesfromtheperspectiveoftheuser.Inthefirststage,DataCollection,individualsitescollecteddatausedtomakesimplecalculationssuchassummationsoraverages.Informationgeneratedatthisstepansweredbusinessquestionsrelatedtofiguresderivedfromdatacollectionsites,suchastotalrevenueoraveragetotalrevenueoveraperiodoftime.SpecificapplicationprogramswerecreatedforcollectingdataandcalculationsThesecondstep,DataAccess,useddatabasestostoredatainastructuredformat.Atthisstage,company-widepoliciesfordatacollectionandreportingofmanagementinformationwereestablished.Becauseeverybusinessunitconformedtospecificrequirementsorformats,businessescouldquerytheinformationsystemregardingbranchsalesduringanyspecifiedtimeperiod.Onceindividualfigureswereknown,questionsthatprobedtheperformanceofaggregatedsitescouldbeasked.Forexample,regionalsalesforaspecifiedperiodcouldbecalculated.Thankstomulti-dimensionaldatabases,abusinesscouldobtaineitheraglobalviewordrilldowntoaparticularsiteforcomparisonswithitspeersDataNavigation.Finally,on-lineanalytictoolsprovidedreal-timefeedbackandinformationexchangewithcollaboratingbusinessunitsDataMining.Thiscapabilityisusefulwhensalesrepresentativesorcustomerservicepersonsneedtoretrievecustomerinformationon-lineandrespondtoquestionsonareal-timebasis.Informationsystemscanquerypastdatauptoandincludingthecurrentlevelofbusiness.Oftenbusinessesneedtomakestrategicdecisionsorimplementnewpoliciesthatbetterservetheircustomers.Forexample,grocerystoresredesigntheirlayouttopromotemoreimpulsepurchasing.Telephonecompaniesestablishnewpricestructurestoenticecustomersintoplacingmorecalls.Bothtasksrequireanunderstandingofpastcustomerconsumptionbehaviordatainordertoidentifypatternsformakingthosestrategicdecisions?anddataminingisparticularlysuitedtothispurpose.Withtheapplicationofadvancedalgorithms,datamininguncoversknowledgeinavastamountofdataandpointsoutpossiblerelationshipsamongthedata.Datamininghelpbusinessesaddressquestionssuchas,“WhatislikelytohappentoBostonunitsalesnextmonth,andwhy?”Eachofthefourstageswererevolutionarybecausetheyallowednewbusinessquestionstobeansweredaccuratelyandquickly[4].Thecorecomponentsofdataminingtechnologyhavebeendevelopingfordecadesinresearchareassuchasstatistics,artificialintelligence,andmachinelearning.Today,thesetechnologiesaremature,andwhencoupledwithrelationaldatabasesystemsandacultureofdataintegration,theycreateabusinessenvironmentthatcancapitalizeonknowledgeformerlyburiedwithinthesystems.ApplicationsofdataminingDataminingtoolstakedataandconstructarepresentationofrealityintheformofamodel.Theresultingmodeldescribespatternsandrelationshipspresentinthedata.Fromaprocessorientation,dataminingactivitiesfallintothreegeneralcategoriesseeFig.1:Discovery?theprocessoflookinginadatabasetofindhiddenpatternswithoutapredeterminedideaorhypothesisaboutwhatthepatternsmaybe.PredictiveModeling?theprocessoftakingpatternsdiscoveredfromthedatabaseandusingthemtopredictthefuture.ForensicAnalysis?theprocessofapplyingtheextractedpatternstofindanomalousorunusualdataelements.Dataminingisusedtoconstructsixtypesofmodelsaimedatsolvingbusinessproblems:classification,regression,timeseries,clustering,associationanalysis,andsequencediscovery[3].Thefirsttwo,classificationandregression,areusedtomakepredictions,whileassociationandsequencediscoveryareusedtodescribebehavior.Clusteringcanbeusedforeitherforecastingordescription.Companiesinvariousindustriescangainacompetitiveedgebyminingtheirexpandingdatabasesforvaluable,detailedtransactioninformation.Examplesofsuchusesareprovidedbelow.Eachofthefourapplicationsbelowmakesuseofthefirsttwoactivitiesofdatamining:discoveryandpredictivemodeling.Thediscoveryprocess,whilenotmentionedexplicitlyintheexamplesexceptintheretaildescription,isusedtoidentifycustomersegments.Thisisdonethroughconditionallogic,analysisofaffinitiesandassociations,andtrendsandvariations.Eachoftheapplicationcategoriesdescribedbelowdescribessomesortofpredictivemodeling.Eachbusinessisinterestedinpredictingthebehaviorofitscustomersthroughtheknowledgegainedindatamining[5].2.3.1.RetailThroughtheuseofstore-brandedcreditcardsandpoint-of-salesystems,retailerscankeepdetailedrecordsofeveryshoppingtransaction.Thisenablesthemtobetterunderstandtheirvariouscustomersegments.Someretailapplicationsinclude[5]:Performingbasketanalysis?Alsoknownasaffinityanalysis,basketanalysisrevealswhichitemscustomerstendtopurchasetogether.Thisknowledgecanimprovestocking,storelayoutstrategies,andpromotions.Salesforecasting?Examiningtime-basedpatternshelpsretailersmakestockingdecisions.Ifacustomerpurchasesanitemtoday,whenaretheylikelytopurchaseacomplementaryitem?Databasemarketing?Retailerscandevelopprofilesofcustomerswithcertainbehaviors,forexample,thosewhopurchasedesignerlabelsclothingorthosewhoattendsales.Thisinformationcanbeusedtofocuscost?effectivepromotions.Merchandiseplanningandallocation?Whenretailersaddnewstores,theycanimprovemerchandiseplanningandallocationbyexaminingpatternsinstoreswithsimilardemographiccharacteristics.Retailerscanalsousedataminingtodeterminetheideallayoutforaspecificstore.2.3.2.BankingBankscanutilizeknowledgediscoveryforvariousapplications,including[5]:Cardmarketing?Byidentifyingcustomersegments,cardissuersandacquirerscanimproveprofitabilitywithmoreeffectiveacquisitionandretentionprograms,targetedproductdevelopment,andcustomizedpricing.Cardholderpricingandprofitability?Cardissuerscantakeadvantageofdataminingtechnologytopricetheirproductssoastoimizeprofitandminimizelossofcustomers.Includesrisk-basedpricing.Frauddetection?Fraudisenormouslycostly.Byanalyzingpasttransactionsthatwerelaterdeterminedtobefraudulent,bankscanidentifypatterns.Predictivelife-cyclemanagement?Datamininghelpsbankspredicteachcustomer'slifetimevalueandtoserviceeachsegmentappropriatelyforexample,offeringspecialdealsanddiscounts.譯文：客戶關(guān)系管理中的數(shù)據(jù)挖掘技術(shù)ChrisRygielski,Jyun-ChengWang,DavidC.Yen摘要近年來，技術(shù)的進(jìn)步讓關(guān)系營銷成為一個(gè)現(xiàn)實(shí)。如數(shù)據(jù)倉庫，數(shù)據(jù)挖掘和一系列管理軟件等技術(shù)已經(jīng)取得了客戶關(guān)系管理的新領(lǐng)域，在那里公司可以贏得競爭優(yōu)勢。特別是通過數(shù)據(jù)挖掘中從大型數(shù)據(jù)庫隱藏的預(yù)測信息的提取，企業(yè)可以識(shí)別有價(jià)值的客戶，預(yù)測客戶未來的行為，并使企業(yè)積極進(jìn)取，做出知識(shí)驅(qū)動(dòng)的決策。通過數(shù)據(jù)挖掘移動(dòng)可能超越過去的事件的分析，自動(dòng)化是適應(yīng)于未來的分析，通常用歷史為導(dǎo)向的工具，提供了諸如決策支持系統(tǒng)。數(shù)據(jù)挖掘工具回答了在過去太費(fèi)時(shí)追求的業(yè)務(wù)問題。然而，這些問題的答案使客戶關(guān)系管理成為可能。各種技術(shù)在數(shù)據(jù)挖掘軟件存在，不同類型的應(yīng)用程序都擁有自身的優(yōu)勢和挑戰(zhàn)。在神經(jīng)網(wǎng)絡(luò)和卡方自動(dòng)交互檢測(CHAID)中存在一個(gè)特殊的二分法。雖然不同的方法于大量的境界數(shù)據(jù)挖掘，一些數(shù)據(jù)挖掘類型用于要完成的各項(xiàng)目標(biāo)的使用，對(duì)當(dāng)今的客戶關(guān)系管理理念來說是很必要的。22Elsevier科學(xué)有限公司保留所有權(quán)利。關(guān)鍵詞：客戶關(guān)系管理(CRM)關(guān)系營銷數(shù)據(jù)挖掘神經(jīng)網(wǎng)絡(luò)卡方自動(dòng)交互檢測(CHAID)私隱權(quán)1.簡介今天，一個(gè)新的商業(yè)文化正在發(fā)展。因此,客戶關(guān)系經(jīng)濟(jì)學(xué)在根本途徑中不斷變化，與此同時(shí)企業(yè)都面臨處理這些變化要實(shí)施新的解決方案和戰(zhàn)略的需要。大規(guī)模生產(chǎn)和大規(guī)模營銷的概念，最先是在工業(yè)革命時(shí)被創(chuàng)造，現(xiàn)在正在被新的觀念所取代，其中客戶關(guān)系是中央企業(yè)的問題。今天的企業(yè)越來越通過對(duì)客戶生命周期分析關(guān)注客戶價(jià)值。這些工具和數(shù)據(jù)倉庫技術(shù)，數(shù)據(jù)挖掘和其他客戶關(guān)系管（以客為本的觀點(diǎn)）正在取代由“設(shè)計(jì)?建造?銷售”（以產(chǎn)品為導(dǎo)向的觀點(diǎn)）的舊模式。傳統(tǒng)大規(guī)模營銷的過程在一對(duì)一營銷的新方法上被質(zhì)疑。在傳統(tǒng)的過程中，營銷的目標(biāo)是吸引更多客戶，擴(kuò)大客戶群。不過，考慮到獲取新客戶的成本，它可以更好地進(jìn)行與現(xiàn)有客戶的業(yè)務(wù)。在這樣做時(shí)，營銷重點(diǎn)從客戶群的寬度轉(zhuǎn)移到每個(gè)客戶的深度需求。該性能指標(biāo)從市場份額到所謂的“錢包份額”變化。企業(yè)不只是為了進(jìn)行交易而應(yīng)付客戶，他們把握了運(yùn)用服務(wù)體驗(yàn)銷售產(chǎn)品，并努力與每一位客戶建立長期合作關(guān)系的機(jī)會(huì)?；ヂ?lián)網(wǎng)的出現(xiàn)，無疑有助于市場重點(diǎn)的轉(zhuǎn)變。由于網(wǎng)上信息變得更方便和豐富，消費(fèi)者變得更加明智和成熟。他們從所有正在提供的信息中知道，他們要求最好的。為了