基于分形的數(shù)據(jù)流聚類算法研究

1、合肥工業(yè)大學(xué)碩士學(xué)位論文基于分形的數(shù)據(jù)流聚類算法研究姓名：羅義欽申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：管理科學(xué)與工程指導(dǎo)教師：張公讓倪志偉20090401ResearchontheAlgorithmofFractalClusteringDataStreamAbstractDatastreamconsistsofaseriesofordinalcoming，boundless，dynamicdataRealworldapplicationsofteng

2、eneratehugeamountofdatastreamside，whichchallengesefficientprocessingandminingduetoitsspecialcharacteristicsThedisadvantagesoftraditionaldataminingalgorithmsinminingdatastreamisindicatedbymanyresearchers，meanwhile，thesedi

3、sadvantagesalsopromotetheresearchesofimprovingofexistingdataminingalgorithmsandcreatingnewdata。streamminingalgorithmsAsaimportantproblemindatastreammining，clusteringtechniqueemployedintheseapplicationsshouldbeeffectivein

4、termsofspaceusageThishasreceivedconsiderableattentioninthepastfewyearsduetoitsresearchvalueandincreasingamountofimportantanceinnumerousapplicationsInthethesis，fractalandsomeclassicalalgorithmsforclusteringdatastreamhaveb

5、eensystematicallystudiedandcomprehensivelysummarizedOnthebasicofpreviousresearch，thenovelalgorithmsforfractalclusteringdatastreamareproposedFirstly，thethesisintroducessomebasicknowledgeofdataminingandfractalandclustering

6、andsomeclassicalalgorithmsforclusteringdatastreamSecondlythethesispresentanalgorithmwhichisbasedonfractaltoclusterdatastreamandusesthechangeoffractaldimensiontomeasuretheselfsimilaritybetweendataandclustersWithnoisycondi

7、tion，thealgorithmcandiscoverarbitraryshapeclustersthatreflectthenaturalgroupstatusofdatastreamTheexperimentsshowthegoodperformanceandeffectivityofFClustreamFinallythethesisconductacomprehensivesummaryanddiscussionofFclus