1 / 41
文档名称:

数据挖掘5.pptx

格式:pptx   大小:1,397KB   页数:41页
下载后只包含 1 个 PPTX 格式的文档,没有任何的图纸或源代码,查看文件列表

如果您已付费下载过本站文档,您可以点这里二次下载

分享

预览

数据挖掘5.pptx

上传人:wz_198613 2020/3/1 文件大小:1.36 MB

下载得到文件列表

数据挖掘5.pptx

文档介绍

文档介绍:2020/3/putationandDataGeneralization2020/3/12Whatisconceptdescription?parisonofthedatathesimplestkindofdescriptivedataminingsometimescalledclassdescriptionwhentheconcepttobedescribedreferstoaclassofobjectsCharacterization:parison(discrimination):paringtwoormorecollectionsofdata2020/3/13DatageneralizationbothcharacterizationanddiscriminationarebasedondatageneralizationandsummarizationDatageneralizationaprocesswhichabstractsalargesetoftask-relevantdatainadatabasefromarelativelylowconceptualleveltohigherconceptuallevelsDatageneralizationapproaches:datacubeapproachattribute-orientedinductionapproach2020/3/14DatacubeapproachThedataforanalysisarestoredinamultidimensionaldatabase,ordatacubegeneralizationandspecializationcanbeperformedonadatacubebyroll-upanddrill-downthisisnotanapproachforconceptdescription,onlyfordatageneralizationLimitations:hetypesofdimensionstosimplenonnumericdataandofmeasurestosimpleaggregatednumericvaluesconcepthierarchiescanbeautomaticallygeneratedfromnumericdatatoformnumericdimensions,however,mercialsystemscannottellwhichdimensionsshouldbeusedandwhatlevelsshouldthegeneralizationreach2020/3/15IsOLAPenough?OLAPrestrictedtocertainkindsofattributesandmeasuretypesuser-plexdatatypesoftheattributesandtheiraggregationsamoreautomatedprocess2020/3/16Attribute-,,,KDDWorkshopatIJCAI-89initsinitialproposal,AOIisarelationaldatabasequery-oriented,generalization-based,putationcanalsobeusedcanbeusedforbothcharacterizationanddiscriminationgeneralidea:collectthetask-relevantdataperformgeneralizationbyattributeremovalorattributegeneralizationapplyaggregationbymergingidentical,umulatingtheirrespectivecountsinteractivepresentationwithusers2020/3/17SketchofAOIDatafocusingthespecificationoftask-relevantdata,whoseresultistheinitialrelationDatageneralizationattributeremovalifthereisalargesetofdistinctvaluesforanattribute,buteither(1)thereisnogeneralizationoperatorontheattribute,or(2)itshigherlevelconceptsareexpressedinterm