基于免參數(shù)據(jù)挖掘的相異度度量研究.pdf

1、授予單位代碼學(xué)號或申請?zhí)柮芗夃嵷大學(xué)碩士學(xué)位論文1045904301124論文題目：基子兔參數(shù)據(jù)挖掘的相異度度量研究作者姓名：學(xué)科門類：專業(yè)名稱：導(dǎo)師姓名、職稱：文娟娟工學(xué)計算機軟件與理論柴玉梅副教授二零零七年五月ABSTRA(rABSTRACTInrecentyears，dataminingtechniquedevelopeMquicklyTheclusteringandclassificationmethodshavebeenap

2、pliedinmanyfieldsButallthesemethodsneedparameterssetting；theproblemsproducedbyparametersettingattractmanyattentionsfromresearchersInordertesolvetheproblems，theideaofparameterfleedataminingwasproposedbysomeresearchersInth

3、isworkWepresentedthetheoryofdatamininganalyzedtheintluenceofparametcfsettingandindicateditwasonefactsofleadingtheresultofdataminingtofalseSoinordertoavoidtheproblemwemustremovetheparametersfTomeverystepofdataminingproced

4、ureWechoosethedissimilaritymeasureAnewmethodSCDM(s卵吼喊calCompr懿sionBasodDissimilarityMeasure)whichbaseOnKolmogorovcomplexitytheorycoupledwithcompressionwasproposedTheSCDMadaptscompressionalgorithmtoestimateKolmogorovcompl

5、exityBecausecompressionalgorithmistypicallyspaceandtimeefficientSCDMinherittheseadvantagesInthispaperSCDM’SfunctionWaftimplementedusingMATLAB、standardedcompressorandGenCompressonDNAsequenceManyexperimentswereperformedonD

6、NAandtimeseriessequencesandcomparisonofourresultswithotherresultswhichufleEuclideandistanceThenSCDMwasappliedforhierarchicalclusteringAsshownintheexperimentalresultsSCDMdon’trequirethedimensionalityoftwoinstancesbeingcom

7、paredisexactlythesameallowthemissingofsingledatapointIthoods’tsetparameterandwithitshi曲efficiencyitcaneasilyresolvethehi曲dimensionalityinstanceSCDMavoidtheinfluenceofparameterimprovetherightnessofalgorithmandusingitintoh