搜索
摘要me数据挖掘技术在贸易公司客户流失预测中的应用研究作者:周鑫指导教师:研究生毕业论文(申请工程硕士学位)软件学院DataMiningintradingincustomerchurnpredictionAppliedResearchSubmittedinpartialfulfillmentoftherequirementsforthedegreeofMasterofEngineeringSupervisedbyProfessorAssociateProfessorSoftwareInstituteNANJINGUNIVERSITYNanjing,China摘要摘要随着数据挖掘技术的发展，数据挖掘的重要性己经被越来越多的人认可，它是利用己知的数据通过建立数学模型的方法找出隐含的业务规则。在国外很多的行业己经具有成功的应用。例如，贸易行业的应用领域主要有客户关系管理，客户欺诈分析，客户流失分析，客户消费模式分析，市场推广分析等。在国内随着对数据挖掘技术的重视，数据挖掘技术的应用研究也越来越广，其中对贸易行业的客户流失分析就是一大热点。客户流失分析是通过对以往流失客户的历史数据进行分析，找出可能流失客户的特征，及时采取相应的措施，减少客户流失的发生。这对企业降低运营成本，提高经营业绩有着极为重要的意义。本文首先阐述了数据挖掘的定义、功能以及流程，分析了CRM的内涵以及体系框架，并指出了贸易公司客户流失的原因以及实现客户流失预测对贸易公司的必要性，在此基础上将贸易公司客户流失原因归类分析，拟出贸易公司客户流失KPI及相应对策。其次，重点设计了贸易公司的客户流失预测模型。针对常用的决策树分类挖掘算法进行分析，指出了存在的问题，提出了基于加权属性和预剪枝策略的决策树分类挖掘的改进算法。该算法能够更好的解决贸易行业数据挖掘中大数据量，效率高的要求。同时基于决策树挖掘改进算法，对客户流失预测模型的架构进行了总体的设计，对客户相关的数据进行了收集、集成，根据拟定的KPI进行数据重组。并在明确客户流失预测模型构建思想的前提下，给出了基于决策树挖掘改进算法的客户流失预测模型详细的建立过程。接着，针对具体的贸易公司，将本文建立的客户流失预测模型进行应用，详细分析了模型应用过程中所涉及到数据清洗、数据转换以及通过决策树分类挖掘改进算法建立的客户流失预测模型。最后，本文对预测模型的结果进行评估与分析，给出了针对性的对策建议，有效地改善客户流失现象，同时证明了决策树分类改进算法的有效性和实用性。关键词：数据挖掘；客户流失预测；贸易公司；决策树ABSTRACTWithdataminingtechnologydevelopment,theimportanceofdatamininghasbeenrecognizedmoreandmorepeople,itisknowntousethedatathroughtheestablishmentofmathematicalmodelofthemethodtofindhiddenbusinessrules.Manyindustriesinforeigncountrieshasbeenasuccessfulapplication.Forexample,theinsuranceindustryapplicationsaremainlycustomerrelationshipmanagement,customerfraudanalysis,customerchurnanalysis,customerconsumptionpatterns,analysis,marketinganalysis.Inthecountrywiththeimportanceofdatamining,dataminingtechniquesappliedresearchareincreasinglybroad,inwhichtheinsuranceindustry,customerchurnanalysisisahottopic.Customerchurnanalysisisthelossofcustomersthroughthepast,historicaldataanalysistoidentifythecharacteristicsoftheusermaysurrenderpromptlytakeappropriatemeasurestoreducecustomerchurnhappening.Thisenterprisestoreduceoperationalcostsandimprovebusinessperformancehasextremelyimportantsignificance.Thispaperbeginsbydescribingthedefinitionofdatamining,functionandprocesses,analysisofthecontentaswellastheCRMsystemframework,andpointsoutthereasonsforthelossofinsurancecompanyclientsandtheachievementofcustomerchurnpredictionoftheneedforinsurancecompanies,insurancecompanieswillbeonthisbasis,classifythereasonsforcustomerchurnanalysistobeinsurancecompanieslosecustomersKPIandthecorrespondingcountermeasures.Second,thefocusonthedesignoftheinsurancecompany'scustomerchurnpredictionmodels.Decisiontreeforclassificationminingalgorithmscommonlyusedintheanalysis,pointingouttheexistenceoftheissue,basedontheweightedattributesandpre-pruningstrategyforimproveddecisiontreeclassificationalgorithmformining.Thealgorithmcanbeabettersolutiontotheinsuranceindustryindatamininglargedatavolumeandhighefficiencyrequirements.Atthesametimeimprovedalgorithmbasedondecisiontreeminingonchurnpredictionmodelsintheframeworkoftheoveralldesignofthecustomer-relateddatacollection,integration,undertheproposedrestructuringofKPIdata.Andaclearcustomerchurnpredictionmodelbuiltunderthepremiseofthoughtisgiventoimprovethealgorithmbasedondecisiontreeminingcustomerchurnpredictionmodelbuildingprocessindetail.Then,specificcommerciallifeinsurancecompany,willthissetupcustomerchurnpredictionmodelapplication,adetailedanalysisofthemodelapplicationprocessinvolvedinthedatacleansing,dataconversionaswellasthroughimproveddecisiontreeclassificationalgorithmforminingcustomerchurnpredictionmodelestablished.Finally,predictionmodelstoassesstheresultsandanalysisaregivenspecificpolicyproposals,effectivelyimprovecustomerturnover,isalsoshownthatdecisiontreeclassificationalgorithmtoimprovetheeffectivenessandpracticality.Keywords:datamining;customerchurnprediction;insurance;decisiontree目录目录1绪论................................................................................................................................................11.1研究背景及意义............................................................................................................................................11.1.1研究背景...........................................................................................................................11.1.2研究意义..........................................................................................................................22研究意义................................................................................................................................21.2国内外研究现状...........................................................................................................................................31.2.1数据挖掘的研究现状......................................................................................................31.2.2挖掘技术在客户流失预测的研究现状.........................................................................51.3论文研究内容及方法...................................................................................................................................61.4论文的组织结构...................................................................