資源描述：
《機(jī)器學(xué)習(xí)十大算法：樸素貝葉斯》由會員上傳分享，免費在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、Chapter9Na¨?veBayesDavidJ.HandContents9.1Introduction...........................................................1639.2AlgorithmDescription.................................................1649.3PowerDespiteIndependence...........................................1679.4ExtensionsoftheMo
2、del...............................................1699.5SoftwareImplementations..............................................1719.6Examples..............................................................1719.6.1Example1......................................................1719.6.2Exam
3、ple2......................................................1739.7AdvancedTopics.......................................................1749.8Exercises..............................................................175References............................................................
4、......1769.1IntroductionGivenasetofobjects,eachofwhichbelongstoaknownclass,andeachofwhichhasaknownvectorofvariables,ouraimistoconstructarulewhichwillallowustoassignfutureobjectstoaclass,givenonlythevectorsofvariablesdescribingthefutureobjects.Problemsofthiskind,calledproblemsofsuper
5、visedclassi?cation,areubiquitous,andmanymethodsforconstructingsuchruleshavebeendeveloped.Oneveryimportantmethodisthena¨?veBayesmethod—alsocalledidiot’sBayes,simpleBayes,andindependenceBayes.Thismethodisimportantforseveralreasons,includingthefollowing.Itisveryeasytoconstruct,notneedi
6、nganycomplicatediterativeparameterestimationschemes.Thismeansitmaybereadilyappliedtohugedatasets.Itiseasytointerpret,sousersunskilledinclassi?ertechnologycanun-derstandwhyitismakingtheclassi?cationitmakes.And,particularlyimportant,itoftendoessurprisinglywell:Itmaynotbethebestpossibl
7、eclassi?erinanygivenapplication,butitcanusuallybereliedontoberobustandtodoquitewell.Forexample,inanearlyclassicstudycomparingsupervisedclassi?cationmethods,Tit-teringtonetal.(1981)foundthattheindependencemodelyieldedthebestoverallresult,whileManietal.(1997)foundthatthemodelwasmostef
8、fectiveinpredicting
