欢迎来到天天文库
浏览记录
ID:40383624
大小:2.53 MB
页数:88页
时间:2019-08-01
《classification1-full》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、Classification(Part1)1OverviewBasicdecisiontreeclassifier(DT)constructionSometechnicalissuesofDTclassificationEvaluatingclassifiers2ClassificationWhatcommoncharacteristicsaresharedbytheredpeople,andnotbythepurplepeople?test(model)badcredithistorygoodcredithistory3RecordAgeInc
2、omeStudenCredit-Own-idtratingcomputer1<30HighNoBadNo2<30HighNoGoodNo330..HighNoBadYes404>40MediumNoBadYes5>40LowYesBadYes6>40LowYesGoodNo730..LowYesGoodYes408<30MediumNoBadNo9<30LowYesBadYes10>40MediumYesBadYes11<30MediumYesGoodYes1230..MediumNoGoodYes401330..HighYesBadYes4014>4
3、0MediumNoGoodNo4ExamplerulesIfage<30andisnotastudentnotacomputerownerIfage<30andisastudentacomputerownerIfageisbetween30to40acomputerownerIfage>40withagoodcreditratingnotacomputerownerIfage>40withabadcreditratingacomputerowner5DataModeladatasetconsistsofanumberofrecor
4、dseachrecordconsistsofanumberofattributevaluesoneparticularattributeiscalledthelabel(orclass)recordsthatsharethesamelabelvalueformaclasswewanttodiscoverrulesthathelppredict,givenafuture(unclassified)record,whichclasstherecordshouldbelong6SupervisedLearningourgeneralapproach
5、totheclassificationproblemprepareadatasetoflabeledrecordsdrawarandomsample(e.g.,80%),callitthetrainingsetusethetrainingsettotrainaclassifierapplytheclassifiertotherest(e.g.,20%)oftherecords(thetestset)toevaluatetheaccuracyoftheclassifierduringthetrainingphase,theclassifieri
6、sfedwithlabeledrecords(examplesofeachclass),thelearningissupervised.Thismachinelearningapproachiscalledsupervisedlearning.7Decision-TreeClassifier(DT)weconsiderhowtobuildadecision-treeclassifierweassumethatattributesareallnominal–theyonlytakeonafinitesetofnamed-valuesnumerica
7、ttributescanbemappedtonominalattributesbydiscretizationorbinningexample:incomecanbegroupedintolow(<80K),medium(>=80K,<250K),high(>=250K)aperson’sheightcanbeveryshort,short,average,tall,verytall(withasuitablemapping)8Decision-TreeClassifieradecisiontreeisatreestructureprop
8、erties:eachinternalnodedenotesatestonanattribu
此文档下载收益归作者所有
点击更多查看相关文章~~