資源描述：
《基于數(shù)據(jù)挖掘的電信行業(yè)中客戶流失模型的分析與實現(xiàn)》由會員上傳分享，免費在線閱讀，更多相關(guān)內(nèi)容在教育資源-天天文庫。
1、浙江理工大學(xué)碩士學(xué)位論文基于數(shù)據(jù)挖掘的電信行業(yè)中客戶流失模型的研究與實現(xiàn)AbstractInrecentyears，customerchumintelecombecomesserious．ForthethreeoperatorsincludingMobile，ChinaUnicornandTelcomwhoCallassuretheircustomersnottoabandonthedefaultgotoanotherone，atthesametimecangetthelostfromothers，wil
2、lbethefinalwinnerAbittercontesthaslaunchedamongoperatorswhodoalltheycouldtogainmorecustomers．However,operatorspaymuchmoreattentiontocustomerschurnedratherthanthosewhoarereducingtheirconsumption．Hardlyrealizethatthosepeoplearerunningoffgraduallyaspotential
3、lOSScustomers．Telecommunicationsindus仃)，hasamassofdataanddiversitywhichmeansthateverycustomerhasalargenumberofattributeswhichcalledvariablesinthedataminingmodel，suchasARPU，chargeway,downtimesetc．Inordertobettermodel，thispaperdesignswidesheetfromthreeaspec
4、tsofoperatorsendsmessagestoremindcustomer,customerperceptivevalueandvaluebehaviorofcustomerbasedontheassumptionthatthereasonsforcustomerschum．Thendividesthesevariablesintodifferentgroupstodecidewhichvariablestoparticipateinmodelmodeling．Combinedthecustome
5、r’Sowncharacteristicsandthegroupingvariablesthensubdivideallofthecustomers．Customer-chummodelbecomesmoreandmoreintelecommunicationsindustry，inordertoimprovethehitrateofthemodel，thispaperproposesacombinedmodelphilosophy．Thecombinedmodelisbasedontheconstrai
6、ntmodel，predictionmodel，markmodel．Constraintmodelselectsvariableshaslargedistinctiontobeconstraintconditions，predictionmodelscreensvisiblelossandrelativelyobviousvariables，markmodelselectsimplicitcustomerchumvariablestomakeupsmallamountofsamplestoidentify
7、customermorecomprehensively．Eachmodelhasitsownspecialvariables，asaresult，thecombinationmodelproposedinthispaperplaysasignificantroleincustomer-chummodel．ThismodelusesIBMSPSSStatistics，putsforwardtheconceptofpheromonedifferenceinantcolonyalgorithm，usingthe
8、improvedantcolonyalgorithmtosubdividethecustomertoimprovethecustomerclusteringeffect．Canonicaltransformationeliminatetheimpactofthedimensionlesscoefficientstomakethemodelmoreregular．Usinglogisticregressionalgorithmi
