资源描述:
《机器学习十大算法:朴素贝叶斯》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、Chapter9Na¨ıveBayesDavidJ.HandContents9.1Introduction...........................................................1639.2AlgorithmDescription.................................................1649.3PowerDespiteIndependence...........................................1679.4Extensionsof
2、theModel...............................................1699.5SoftwareImplementations..............................................1719.6Examples..............................................................1719.6.1Example1......................................................17
3、19.6.2Example2......................................................1739.7AdvancedTopics.......................................................1749.8Exercises..............................................................175References.............................................
4、.....................1769.1IntroductionGivenasetofobjects,eachofwhichbelongstoaknownclass,andeachofwhichhasaknownvectorofvariables,ouraimistoconstructarulewhichwillallowustoassignfutureobjectstoaclass,givenonlythevectorsofvariablesdescribingthefutureobjects.Problemsofthiskind,c
5、alledproblemsofsupervisedclassification,areubiquitous,andmanymethodsforconstructingsuchruleshavebeendeveloped.Oneveryimportantmethodisthena¨ıveBayesmethod—alsocalledidiot’sBayes,simpleBayes,andindependenceBayes.Thismethodisimportantforseveralreasons,includingthefollowing.Itisver
6、yeasytoconstruct,notneedinganycomplicatediterativeparameterestimationschemes.Thismeansitmaybereadilyappliedtohugedatasets.Itiseasytointerpret,sousersunskilledinclassifiertechnologycanun-derstandwhyitismakingtheclassificationitmakes.And,particularlyimportant,itoftendoessurprisingl
7、ywell:Itmaynotbethebestpossibleclassifierinanygivenapplication,butitcanusuallybereliedontoberobustandtodoquitewell.Forexample,inanearlyclassicstudycomparingsupervisedclassificationmethods,Tit-teringtonetal.(1981)foundthattheindependencemodelyieldedthebestoverallresult,whileManiet
8、al.(1997)foundthatthemodelwasmosteffectiveinpredicting