资源描述：
《决策树算法的研究与改进（research and improvement of decision tree algorithm）》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。
1、决策树算法的研究与改进（Researchandimprovementofdecisiontreealgorithm）Forty-sixthvolumes.FourthJournalofXiamenUniversity(NATURALSCIENCEEDITION)Vol.46..No.4July2007Journal,of,Xiamen,U,niversity(Natural,Science)Jul.2007..ResearchandimprovementofdecisiontreealgorithmFengShaorong*(Schoolofco
2、mputerscienceandengineering,SouthChinaUniversityofTechnology,Guangdong,Guangzhou510641)..Receiveddate:2006..06..01ProjectsupportedbythenaturalsciencefoundationofFujianprovince(A0310008),FujianHiTechResearchInstituteOpenprogramkeyproject(2003H043)funded*workunit:Departmentofco
3、mputerscience,XiamenUniversityEmail:shaorong@,xmu.,edu.,cnAbstract:decisiontreeisanimportantclassificationmethodindatamining.Basedontheresearchandcomparisonofseveralclassicaldecisiontreealgorithms,animproveddecisiontreeisproposedDecisiontreealgorithm:ametricbaseddecisiontree(
4、MBDT).ThisdecisiontreeisactuallyacombinationoflinearclassifieranddecisiontreeThedecisiontreecaneffectivelyreducethenumberoflayersofdecisiontrees,soastoimprovetheclassificationefficiencyofdecisiontree.ThroughtheMBDTclassificationexperiment,thecorrectnessoftheaboveconclusionisv
5、erifiedAndeffectivenessKeywords:decisiontree;metric;ID3algorithm;entropyCLCnumber:TP18...Documentidentificationcode:A....Articlenumber:0438..0479(2007)04..0496..05Thedecisiontree(Decision,T,REE)isthemaincategoryforclassificationandpredictionTechnology,whichlooksatthedecisiont
6、reetablesfromasetofirregularinstancesTheclassificationruleisshownintheformofatop-downrecursiveapproachindecisionmakingTheinternalnodesofthetreecomparetheattributevaluesandjudgetheattributesaccordingtodifferentattributesBreakingdownfromthisnodedownwardleadstotheconclusionofthe
7、leafnodeinthedecisiontreeThis,fromtherootnodetotheleafnode,correspondstoareasonableruleforthewholetreeCorrespondstoasetofexpressionrules,oneofthemostbasedonthedecisiontreealgorithmThegreatadvantageisthatitdoesnotrequiretheusertoknowmuchaboutitduringthelearningprocessKnowledge
8、,aslongasthetrainingcasecanbeexpressedintermsofattributesorconclusio
