欢迎来到天天文库
浏览记录
ID:51732061
大小:862.50 KB
页数:87页
时间:2020-03-30
《《数据仓库与数据挖掘》第9章.ppt》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、第7章:分类和预测Whatisclassification?Whatisprediction?IssuesregardingclassificationandpredictionClassificationbydecisiontreeinductionBayesianClassificationClassificationbyNeuralNetworksClassificationbySupportVectorMachines(SVM)Classificationbasedonconceptsfromassociationr
2、uleminingOtherClassificationMethodsPredictionClassificationaccuracySummary2021/7/211DataMining:ConceptsandTechniquesClassification:predictscategoricalclasslabels(discreteornominal)classifiesdata(constructsamodel)basedonthetrainingsetandthevalues(classlabels)inaclas
3、sifyingattributeandusesitinclassifyingnewdataPrediction:modelscontinuous-valuedfunctions,i.e.,predictsunknownormissingvaluesTypicalApplicationscreditapprovaltargetmarketingmedicaldiagnosistreatmenteffectivenessanalysisClassificationvs.Prediction2021/7/212DataMining
4、:ConceptsandTechniquesClassification—ATwo-StepProcessModelconstruction:describingasetofpredeterminedclassesEachtuple/sampleisassumedtobelongtoapredefinedclass,asdeterminedbytheclasslabelattributeThesetoftuplesusedformodelconstructionistrainingsetThemodelisrepresent
5、edasclassificationrules,decisiontrees,ormathematicalformulaeModelusage:forclassifyingfutureorunknownobjectsEstimateaccuracyofthemodelTheknownlabeloftestsampleiscomparedwiththeclassifiedresultfromthemodelAccuracyrateisthepercentageoftestsetsamplesthatarecorrectlycla
6、ssifiedbythemodelTestsetisindependentoftrainingset,otherwiseover-fittingwilloccurIftheaccuracyisacceptable,usethemodeltoclassifydatatupleswhoseclasslabelsarenotknown2021/7/213DataMining:ConceptsandTechniquesClassificationProcess(1):ModelConstructionTrainingDataClas
7、sificationAlgorithmsIFrank=‘professor’ORyears>6THENtenured=‘yes’Classifier(Model)2021/7/214DataMining:ConceptsandTechniquesClassificationProcess(2):UsetheModelinPredictionClassifierTestingDataUnseenData(Jeff,Professor,4)Tenured?2021/7/215DataMining:ConceptsandTechn
8、iquesSupervisedvs.UnsupervisedLearningSupervisedlearning(classification)Supervision:Thetrainingdata(observations,measurements,etc.)areaccompanied
此文档下载收益归作者所有