资源描述:
《基于多重分形的聚类层次优化算法》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、ISSN1000-9825,CODENRUXUEWE-mail:jos@iscas.ac.cnJournalofSoftware,Vol.19,No.6,June2008,pp.1283−1300http://www.jos.org.cnDOI:10.3724/SP.J.1001.2008.01283Tel/Fax:+86-10-62562563©2008byJournalofSoftware.Allrightsreserved.∗基于多重分形的聚类层次优化算法1,2+12闫光辉,李战怀,党建武1(西北工业大学计算机学院,陕西
2、西安710072)2(兰州交通大学电子与信息工程学院,甘肃兰州730070)FindingNaturalClusterHierarchiesBasedonMultiFractal1,2+12YANGuang-Hui,LIZhan-Huai,DANGJian-Wu1(SchoolofComputerScience,NorthwesternPolytechnicalUniversity,Xi’an710072,China)2(SchoolofInformationandElectricalEngineering,LanzhouJi
3、aotongUniversity,Lanzhou730070,China)+Correspondingauthor:E-mail:yangh@mail.nwpu.edu.cnYanGH,LiZH,DangJW.FindingnaturalclusterhierarchiesbasedonMultiFractal.JournalofSoftware,2008,19(6):1283−1300.http://www.jos.org.cn/1000-9825/19/1283.htmAbstract:Aclusterisacollect
4、ionofdataobjectsthataresimilartooneanotherwithinthesameclusterandaredissimilartotheobjectsinotherclusters.Moreover,therewillexistmoreorlesssimilaritiesamongtheselargeamountsofinitialclusterresultsinreallifedataset.Accordingly,analyzermayhavedifficultytoimplementfurt
5、heranalysisiftheyknownothingaboutthesesimilarities.Therefore,itisveryvaluabletoanalyzethesesimilaritiesandconstructthehierarchystructuresoftheinitialclusters.Thetraditionalclustermethodsareunfitforthisclusterpost-processingproblemfortheirfavoroffindingtheconvexclust
6、erresult,impracticalhypothesisandmultiplescansofthedataset.Basedonmultifractaltheory,thispaperproposestheFCHO(fractal-basedclusterhierarchyoptimization)algorithm,whichintegratestheclustersimilaritywithclustershapeandclusterdistributiontoconstructtheclusterhierarchyt
7、reefromthedisjointinitialclusters.Theelementarytime-spacecomplexityoftheFCHOalgorithmispresented.SeveralcomparativeexperimentsusingsyntheticandreallifedatasetshowtheperformanceandtheeffectivityofFCHO.Keywords:datamining;clustering;multifractal;post-processing;optimi
8、zation摘要:大量初始聚类结果之间存在强弱不同的相似性,会给用户理解与描述聚类结果带来不利影响,进而阻碍数据挖掘后续工作的顺利展开.传统聚类算法由于注重聚类形状及空间邻接性,或者考虑全局数据分布密度的均匀性,实际中均难以解决这一类问题.为此,提出了基于分形的聚类层次优化算