數(shù)據(jù)挖掘introduction管理學(xué)教授財(cái)經(jīng)信息管理與工程學(xué)院

Data—Data-Mining Prof.Dr.Prof.Dr.LIUPagePageProf.PageProf.Dr.LIUPageProf.PageProf.Dr.LIU

概率 Prof.Dr.Prof.Dr.LIUPage Prof.Dr.Prof.Dr.LIUPagePagePageProf.Prof.Dr.LIUProf.Dr.Prof.Dr.LIUPage(First-principlemodels–0

TheDataTotalnewdisk(TB)since Grossman,C.informationoverloadbehappywithlargedataProf.Dr.Prof.Dr.LIUPageProf.PageProf.Dr.LIU“NecessityistheMotherof對(duì)大型的、復(fù)雜的、信息豐富的數(shù)據(jù)集的理解是幾乎所有運(yùn)用基于計(jì)算機(jī)的方法（包括新技術(shù)）從數(shù)據(jù)中獲得有用PageProf.PageProf.Dr.LIUTurningaproduct-focusedorganizationintoacustomer-centriconetakesmorethandataminingPageProf.PageProf.Dr.LIU產(chǎn)品SAS(EnterpriseMinerIBMIntelligentMiner,SPSS(Clementine),Megaputer(PolyAnalyst),……DataminingisnotjustaboutproducingtechnicalEnergyKnowledgeCreativityprofessionalcreativethinkingawillingnesstoseeproblemsinadifferent可伸縮性高維性（High異構(gòu)數(shù)據(jù)和復(fù)雜數(shù)據(jù)（HeterogeneousandComplex數(shù)據(jù)所有權(quán)與分布（DataOwnershipand

Prof.Dr.Prof.Dr.LIUPageP(是)=P(是|紅球P(紅球+P(是|白球P(白球Garbagein,garbagePageProf.Dr.PageProf.Dr.LIU模型集（modelThemodelcanonlybeasgoodasthedatausedtocreateit.代表了過去用來吸引客戶的信息、優(yōu)惠以及

Prof.Dr.Prof.Dr.LIUPageProf.Dr.LIUProf.Dr.LIUPageAstudyofcustomerattrition客戶流失mayshowthatastrongpredictorofcustomerleavingisthewaytheywereacquired(獲取).Itistoolatetogobackandchangethatforexistingcustomers,butthatdoesnotmaketheinformationuselessFutureattritioncanbereducedbychangingthemixofacquisitionchannelstofavorthosethatbringinlonger-lastingcustomers.aniterativeprogressisdefinedbythrougheitherautomaticormanualmostusefulinanexploratoryanalysisnopredeterminednotionsaboutwhatwillconstitutean“interesting”outcome.Dataminingisthesearchfornew,valuable,andnontrivialinformationinlargevolumesofdata.BestresultsareachievedbytheknowledgeofhumanexpertsindescribingproblemsandgoalsthesearchcapabilitiesofProf.Dr.Prof.Dr.LIUPage定義：數(shù)據(jù)挖掘是從已知數(shù)據(jù)集合中發(fā)現(xiàn)各種模型、概（DataMiningisaprocessofdiscoveringvariousmodels,summaries,andderivedvaluesfromagivencollectionofdata.）DataminingisaniterativeThegeneralexperimentalprocedureadaptedtodata-miningproblemsinvolvesthefollowingsteps:StatetheproblemandformulatetheDomain-specificknowledgeandexperienceareusuallyProf.Dr.Prof.Dr.LIUPageCollecttheTherearetwodistinctDesignedObservationalapproach.–randomdataPreprocessingtheTwocommonOutlierdetection(andScaling,encoding,andselectingtwodatapreparationEstimatethetheselectionandimplementationoftheappropriatedata-miningtechnique.InterpretthemodelanddrawData-miningmodelsshouldhelpindecisionModelsneedtobeinterpretableinordertobeAgoodunderstandingofthewholeprocessisimportantforanysuccessfulDataMiningandBusinesstosupportbusiness DataDataInformation

EndDataStatisticalAnalysis,QueryingandDataWarehouses/DataDataPaper,Files,InformationProviders,DatabaseSystems,

Prof.Dr.Prof.Dr.LIUPage結(jié)構(gòu)化數(shù)據(jù)（structuredwell-definedfieldswithnumericoralphanumeric半結(jié)構(gòu)化數(shù)據(jù)（semi-structuredelectronicimagesofbusinessdocuments,medicalreports,executivesummaries,andrepairmanualsThemajorityofwebavideorecordedbyasurveillancecamerainadepartment傳統(tǒng)型數(shù)據(jù)（traditionalstructuredMostofthecurrentdataminingmethodsandcommercialtoolsareappliedto非傳統(tǒng)型數(shù)據(jù)：semi-structuredandunstructurednontraditional(alsocalledmultimediathedevelopmentofdata-miningtoolsisprogressingatarapidratethestandardmodelofstructuredacollectionofinatabularform,orintheformofasinglePagePageProf.Dr.LIUindependentvariables(inputvariables)–dependentvariables(outputvariables)–additionalvariables(unobservedvariables)–influencesystembehavior,butthecorrespondingvaluesarenotThehighcomplexityandthecostofmeasurementsforthesenotundersta