大型超市客户特征分析及其营销策略

上海交通大学硕士学位论文大型超市客户特征分析及其营销策略姓名：徐海兵申请学位级别：硕士专业：企业管理指导教师：黄沛20050101THECHARACTERISTICANALYSISOFHYPERMARKETCUSTOMERSANDMARKETINGSTRATEGYABSTRACTAtpresent,Chineseretailindustryhasbeenopenforforeigncapitalcompletely.Theopenmeansthatforeigncapitalcanengageincommoditycirculationwithmanyways,andalsomeansthatChinawillenterintothetideofglobalizationfurther.PolarizationhasappearedinChineseretailindustry.Themarketshareofdepartmentstoreisbecomingsmallerandsmaller,andthesupermarkethasbecomethemainretailform,andhypermarketachievesthegreatestdevelopment.Modernmarketingtheoryrequiresthatthecorporationsshouldfocusontheircustomers.Hypermarketcorporationsarethelastpartofthewholecommoditycirculation,andfacealotofcomplexcustomer.Thecompetitionamongdifferenthypermarketcorporationsmainlyfocusesoncustomerresources.Whoknowscustomermore,whowillbecometheultimatewinner.Thus,hypermarketcorporationsfacestheimportantsubjectthathowtheycanminethecharacteristicsoftheircustomers.Andthisthesisalsofocusesonthesubject.WiththebackgroundofChineseretailindustry’sopentoforeigncapital,thisthesispresentsthesubjectofanalyzingcustomercharacteristics.Withdataminingtechniquesasthetool,thisthesisputsforwardacustomer-characteristic-analysismodel.Withthismodel,weanalyzeallthetradedatawiththeassistanceofSQLServerandSASEnterpriseMiner.Thenweobtainthecharacteristicsofhigh-valuecustomersandhigh-potentialcustomers.Intheend,thisthesisevaluatesthecharacteristicrulesanddiscusseshowthesecharacteristicrulescanbeappliedintohypermarketcorporations’marketing.Thehighlightsofthisthesisarelistedasfollowing:1)Analyzingwiththeactualtradedataofthehypermarket,andmanyoldresearchesfocusingonqualitativeanalysis.2)Usingfive-interzonemethodwhenlookingfordifferentcustomers,andfiveinterzonemethodisatransformationof80/20rule.3)Puttingforwardacustomer-characteristic-analysismodel,andanalyzingwithdecisiontreemethod.Atthesametime,comparingdifferentmodels,andselectingthebestmodel.4Analyzingmarketingstrategyofhypermarket,basedoncustomercharacteristicsandtheoryofconsumerbehaviortheory.KEYWORDS:Hypermarket,DataMining,DecisionTree,MarketingStrategy2005118200511820051182005118313200511811.1200412111-11-1041211493000303000808080901993GDPGDPGDP19932003GDP34560.5200311669412462.1200345842200422524951-11-11-120033908920905.3%1349434.5%1541039.4%31057.9%444711.4%5431.4%52.5%CTR20042000200424%53%41%22%200331-22003461-22003200387.30%32.35%88.50%29.60%11.212.70%21%11.50%33%0.4541.21.2.11.2.251.362.120802-12.2DataMiningKDD19701985SQL1976(SQL)1970198019611956AIWeb2000199019601996DB22-172-22.32.3.112KDD2-28KMeans3()R.AgrawalAprioriApriori2.3.21TrainingSetRecordClassLabel12(,,...,;)nvvvcivck-29CART3772.42.4.1(DecisionTree)2-3CLS(ConceptLearningSystem)()(categorical)(continuous)1032-12-3315000A2-118000A10000B20000A19000B165000A1()2()3NP-hardsplitattributesgoodnessfunction(InformationGain)ID3C4.5A150002-311(GiniIndex)CARTSLIQSPRINTIntelligentMiner(IBM)2c(Chi-squaredstatistic)CHAID(Relevance)G(GStatistics)1InformationGainID3QuinlanC4.5ID3ID3Entropy12×××nEFFF=LnjFE12,,,neVVV=LjjVF∈1,2,,jn=LPENEE2EPENEpnID31EE2(2-1)AAV12{,,,}vVVVLEV12{,,,}vEEELiEiPiNiE(,)iiIPNA()(,)viiiiiPNEAIPNPN+=+∑(2-2)A()(,)()gainAIpnEA=-(2-3)ID3()gainA(()EA)AAEViEA12,,,vBBBLID3ID3ID3(1)ID3(2)(3)ID322(,)loglogppnnIpnpnpnpnpn=--++++12(4)ID3QuinlanC4.5C4.5ID32(GiniIndex)Breiman1984ClassificationandRegressionTreesTNGini21()1[(|)]NjGinitpjt==-∑(2-4)(/)pjtjt()Ginit0()GinitkGini1()()kisplitinGiniTGiniin==∑(2-5)kininPsplitGini32cCHAIDChi-SquareAutomaticInteractionDetection2c2c132.4.21943PittsMcCullochM-P1949HebbHebbHebb1961RosenblattPerceptron1982HopfieldHopfield1984HopfieldAI2-4J1{}jJjx=1(,,)jjjTNNRxxx=∈L11{}jJjOR=⊂11():gxRR→1(,,)TNN=∈L1Rq∈1()()NjjjjnnnOgWgWzxqxq===-=-∑g1,,jJ=L(2-6)jzWNR14()gxJ1{}jJjx=1(,,)jjjTNNRxxx=∈L11{}jJjOR=⊂11():gxRR→1(,,)TNN=∈L1Rq∈1()()NjjjjnnnOgWgWzxqxq===-=-∑g1,,jJ=L(2-7)jzWNR()gxBPHopfieldKohonenARTz2-41x2x11Nx+=-NxW1W2qWN()gg15(1)(2)()2.4.32.4.4RoughSetZ.Pawlak1982()16()()MarzenaK()()()()NP-hard()()(1)Holte(onerulediscretizer)(2)(3)MohuaBanerjee()(1)(2)17(3)(4)2.4.5()()()3()fxfitness()fx(1)(2)(3)cP(4)cP()(1)(2)18(3)(4).Muhienbein(5)2.4.6k-(case-basedreasoning)2.52-52-52.6SQLServer2000SASSAS/EMEnterpriseMiner19SQLServer2000WindowsSASSAS/EM203.118522030205090Internet200489(GB/T18106-2004)171Convenience1002Supermarket60003Hypermarket60004CashandcarryWarehouse600040005Specialty216Department6000-200007ShoppingMall552030083.23.2.11993GDPGDPGDPGDP3.2.22000200340%20009822003357520033908920905.3%1349434.5%1541039.4%31057.9%444711.4%52.5%5.3%2267.1%3.2.3CTR20042000200424%53%41%22%3.2.41232343.31.2.3.4.5.3-13-2909024GDP(1)3-13-225(2)(3)(4)(5)2000CR31%200126,274.1BrachmanAnandReinartzWilliamsFayyad4-1CRISP-DM4-24-1Fayyad/28Fayyad1.2.3.4.5.4-1Fayyadnn-11n-1Fayyad1.FayyadFayyad2.Fayyad4-2CRISP-DM29CRISP-DMCross-IndustryProcessforDataMiningCRISP-DMFayyadCRISP-DMFayyad4-31.2.3.4.5.6.7.4-3/304.24.3314.4200200002020004.52-440%4-41.NULLs32348788022.4.6Hash4-4334.7SASEnterpriseMiner4.1SASE/MWindowsXPIntel(InformationGain)(GiniIn