资源描述:
《数据仓库与数据挖掘(2).ppt》由会员上传分享,免费在线阅读,更多相关内容在工程资料-天天文库。
1、Chapter20:DataAnalysisChapter20:DataAnalysisDecisionSupportSystemsDataWarehousingDataMiningClassificationAssociationRulesClusteringDecisionSupportSystemsDecision-supportsystemsareusedtomakebusinessdecisions,oftenbasedondatacollectedbyon-linetransaction-processingsystems.Examplesofbusinessdecision
2、s:Whatitemstostock?Whatinsurancepremiumtochange?Towhomtosendadvertisements?ExamplesofdatausedformakingdecisionsRetailsalestransactiondetailsCustomerprofiles(income,age,gender,etc.)Decision-SupportSystems:OverviewDataanalysistasksaresimplifiedbyspecializedtoolsandSQLextensionsExampletasksForeachpr
3、oductcategoryandeachregion,whatwerethetotalsalesinthelastquarterandhowdotheycomparewiththesamequarterlastyearAsabove,foreachproductcategoryandeachcustomercategoryStatisticalanalysispackages(e.g.,:S++)canbeinterfacedwithdatabasesStatisticalanalysisisalargefield,butnotcoveredhereDataminingseekstodi
4、scoverknowledgeautomaticallyintheformofstatisticalrulesandpatternsfromlargedatabases.Adatawarehousearchivesinformationgatheredfrommultiplesources,andstoresitunderaunifiedschema,atasinglesite.Importantforlargebusinessesthatgeneratedatafrommultipledivisions,possiblyatmultiplesitesDatamayalsobepurch
5、asedexternallyDataWarehousingDatasourcesoftenstoreonlycurrentdata,nothistoricaldataCorporatedecisionmakingrequiresaunifiedviewofallorganizationaldata,includinghistoricaldataAdatawarehouseisarepository(archive)ofinformationgatheredfrommultiplesources,storedunderaunifiedschema,atasinglesiteGreatlys
6、implifiesquerying,permitsstudyofhistoricaltrendsShiftsdecisionsupportqueryloadawayfromtransactionprocessingsystemsDataWarehousingDesignIssuesWhenandhowtogatherdataSourcedrivenarchitecture:datasourcestransmitnewinformationtowarehouse,eithercontinuouslyorperiodically(e.g.,atnight)Destinationdrivena
7、rchitecture:warehouseperiodicallyrequestsnewinformationfromdatasourcesKeepingwarehouseexactlysynchronizedwithdatasources(e.g.,usingtwo-phasecommit)istooexpensiveUsuallyOKtohaveslightlyout-of-datedataatwarehouseData/upd