基于最小相关和最大依赖度准则的属性约简研究

《基于最小相关和最大依赖度准则的属性约简研究》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、ClassifiedIndex:CODE:10075U.D.C:NO:20101200ADissertationfortheDegreeofM.EngineeringResearchonAttributeReductionBasedonthePrincipleofMinimalMutualInformationandMaximalDependencyCandidate:WanLiyanSupervisor:Prof.ZhaiJunhaiProf.WangXizhaoAcademicDegreeAppliedfor:Mast

2、erofEngineeringSpecialty:ComputerAppliedTechnologyUniversity:HebeiUniversityDateofOralExamination:May,2013摘要摘要粗糙集作为一种新的处理不精确、不确定和不完备数据的数学工具，已被广泛应用于模式识别、数据挖掘和决策分析等众多领域。粗糙集的主要应用是属性约简，属性约简在利用粗糙集技术进行分类规则提取过程中发挥着关键作用。目前存在的属性约简的方法，都没有考虑属性相关。基于属性依赖度的属性约简算法作为广泛应用的属性约简方法之一

3、，仅仅选择相对于决策属性依赖度大的条件属性。本文指出仅仅选择依赖度大的条件属性对于进行有效属性约简是不够的，还应考虑属性冗余即条件属性间的相关性。本文提出了一种新的属性约简方法——基于最小相关和最大依赖度准则的属性约简，选择依赖度大相关性小的属性构造约简，研究了离散值决策表和连续值决策表的属性约简计算问题，并在UCI数据集上进行了实验，实验结果验证了本文方法的可行性和有效性，尤其是能够提高分类精度。关键词粗糙集属性约简属性冗余最大依赖度最小相关IAbstractAbstractAsanewmathematicaltool,

4、RoughSet(RS)canefficientlydealwiththeproblemswithvagueness,uncertaintyandincompleteness,andRShasbeenwidelyappliedtomanyfields,suchaspatternrecognition,datamining,decisionanalysisetal.ThemainapplicationofRSistoconductattributereductionwhichplayskeyroleintheprocesso

5、fextractingclassificationruleswithroughsettechnique.Uptonowtheexistingattributereductionmethodsdonotconsiderattributerelevance.Asoneofthemethodsextensivelystudiedforfindingareduct,methodbasedonattributedependencyonlyselectstheattributeshavingmaximaldependencyanddo

6、notconsiderthemutualrelevanceamongtheattributesinthereduct.Thispaperproposesanovelattributereductionmethodbasedontheprincipleofminimalrelevanceandmaximaldependencywhichdecouplesmaximaldependencyaboutdecisionattributeandminimalrelevanceaboutconditionattributes.Thec

7、omputationofattributereductsofdecisiontablewithdiscretevalueandcontinuousvaluesareinvestigated,andtheexperimentsonUCIdatasetsareconducted.Theexperimentalresultsshowthatourproposedmethodisfeasibleandeffective.Especially,itcanimproveclassificationaccuracy.KeywordsRo

8、ughsetAttributereductionAttributeredundancyMaximaldependencyMinimalrelevanceII目录目录第1章绪论.................................................................