基于粗糙集的属性约简算法及其应用研究

《基于粗糙集的属性约简算法及其应用研究》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、江南大学硕士学位论文基于粗糙集的属性约简算法及其应用研究姓名：颜艳申请学位级别：硕士专业：检测技术与自动化装置指导教师：杨慧中20080301AbstractRoughSet限S)theory,introducedbyPawlakZ，isanovelmathematicaltooltodeal埘tllvaguenessanduncertainty．Itisapowerfulmathematicaltoolforanalyzinguncertain，fuzzyknowledgeandcaneffectivelydeal谢t11theimpre

2、cise，incomplete，oruncertaindata．Nowithasattractedmuchattentionofresearchersaroundtheword．Inrecentyears，ithasbeensuccessfullyappliedtodatamining，machinelearning，knowledgediscoveryfromdatabase，decisionsupportsystems，faultdiagnosisetc．Thisarticleemphaticallystudiesononeofthei

3、mportantproblemofRoughSettheory—也ereductionofthedecisiontable．Attributereductionpreservestheoriginalmeaningandreducestheirrelevantandunimportantknowledge．Thedetailsarestudiedasfollows：Inregardtoacompleteanddiscreteinformationsystem，considerattributereductionintheviewofinfo

4、rmationtheory．Adevelopedattributeimportancemeasuremethodisdefinedbasedonthemutualinformationbetweenselectedattributeanddecisionattribute，andthemeasureisusedastheheuristicinformationintheproposedalgorithm．Conditionalinformationentropyisusedtocomputerelevanceofattributesandi

5、tisusedinfitnessfunctionofgeneticalgorithmtoassurereductionhasfewattributesandrelevancebetweenattributes．TraditionalRoughSettheoryisgenerallyincapableofhandlingincompleteinformationsystem．AfterstudyingtheextensionsofRoughSetmodel，pointouttheirshortages．Foressentialityofatt

6、ributeexistingdifference，adevelopedattributeimportancemeasuremethodisdefinedbasedonthedifferencedegreeofattributes。It’Sproposed趴attributereductionalgorithmbasedonconnectiondegreeofessentialityofattribute．Anexampleshowsthattheproposedalgorithmisaneffectivemethod．AnothertheR

7、oughSettheorydefectwhichblocksitsdevelopmentandapplicationisthatitcallnotbeemployedoncontinuousvaluesdirectly．Previouslydiscretizationmethodisappliedbeforehandinordertotransformthedataintodiscretevalues，butthismayresultininformationloss．nlenotionsofsimilaritybetweenobjects

8、andimprovedgeneralimportantdegreeofanattributeareintroduced．Theglobalsimilaritymeasurebet