文档介绍：———————_——_———————————一———-—————————————————————————————————————_—————————————_———————————————————————一 ABSTRACT Decision treeclassificationlearning algorithm is one ofthe most widely usedand very practical inductive inference isof much theoretical and practical significance intheartificialintelligence kingdom such asmachine learning and data miningIn themany decision treelearning algorithms,the most influential one istheID3, which takes thedescending velocity of the information entropy as test attribute selection criterion However,as iswellknown,ID3 has theshortage such aslearning logicalexpressions andleaning totheattribute which takesmore on ID3 algorithm,this thesisattempts toimpove Oillearning logicalexpressions. We firstintroduce extensive matrix theory inlearning from examples and the optimization problem indecision tree learning,the information theory principle and theimplementation ofID3 algorithm and thepruning principle Then,aiming attheID3’Sdefect on learninglogicalexpressions,we pull forward a decision treesimplificationalgorithm based on inclusion rule(DTSA-BOIR,abbr., BOIR、to simplify thedecision treeconstructed witllII) traverses each node ofthe ID3 decision pares itssubtreesand,ifthe root attributes ofeachsubtreeare thesame andsome corresponding branches of all thesubtrees are identical,changes thehierarchical relationship ofthecorrelative attributes inthe decision treeandmerges the identicalbranches respectively. Thisthesisimplements thealgorithm BOIR forlearning logicalexpressions and tests BOIR wittlsome datasets inthe FAMn family,and the data got from the experiment validatesthevalidity ofthealgorithm. Key words:learning fromexamples,decision tree,information entropy, simplifying decision tree,parison,merging branches Il 合肥工业大学本论文经答辩委员会全体委员审查,确认符合合肥工业大学硕士学位论文质量要求。答辩委员会签名主席:辱%铭似生就缛琴旋委员: 导师:‰孚:戡 r, 独创性声明本人声明所呈交的学位论文是本人在导师指导下进行的研究工作及取得的研究成果。据我所知,除了文中特,BiJDH以标注和致谢的地方外,论文中不包含其他人己经发表或撰写过的研究成果,也不包含为获得合肥工业大学或