资源描述：
《一种基于主存_tree的高维数据自相似连接处理》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、计算机研究与发展ISSN1000-1239PCN11-1777PTPJournalofComputerResearchandDevelopment46(6):995-1002,2009一种基于主存$-tree的高维数据自相似连接处理1,31,2刘艳郝忠孝1(哈尔滨理工大学计算机与控制学院哈尔滨150080)2(哈尔滨工业大学计算机科学与技术学院哈尔滨150001)3(长春大学计算机科学技术学院长春130022)(liuyan6374@yahoo.com.cn)A$-TreeBasedSimilarityJoinProc
2、essingforHigh-DimensionalData1,31,2LiuYan,andHaoZhongxiao1(CollegeofComputerandControl,HarbinUniversityofScienceandTechnology,Harbin150080)2(CollegeofComputerScienceandTechnology,HarbinInstituteofTechnology,Harbin150001)3(CollegeofComputerScienceandTechnology,Ch
3、angchunUniversity,Changchun130022)AbstractThesimilarityjoin,animportantdataminingprimitive,canbesuccessfullyappliedtospeedingupapplicationssuchassimilaritysearch,dataanalysisanddatamining.Sofarmostofresearchesfocusontheexecutionofhigh-dimensionaljoinsoverlargeam
4、ountsofdiskbaseddata.Theincreasingsizesofmainmemoryavailableoncurrentcomputers,andtheneedforefficientprocessingofspatialjoinssuggestthatspatialjoinsforalargeclassofproblemscanbeprocessedinmainmemory.$-treeisanovelmult-ilevelindexstructure,itcanspeedupthehigh-dim
5、ensionalqueryinmainmemoryenvironmentandhasbeenproventobeanefficientindexmethod.Eachlevelinthe$-treerepresentsthedataspaceatdifferentdimensionalities:thenumberofdimensionsincreasestowardstheleaflevelwhichcontainsthedataattheirfulldimensions.Theremainingdimensions
6、areobtainedusingprincipalcomponentanalysis.Usingthepropertiesof$-tree,asimilarityjoinalgorithmonthebasisofindexstructure$-tree,$-tree-join,ispresented.Thetop-downschemecanusefewernumberofdimensions,computethedistancesandefficientlycompletejoinprocessing.Experime
7、ntalresultsindicatethat$-tree-joinoutperformsthestate-of-the-artalgorithm,EGO-join,*andEGO-joinbyawidemargin,andisanefficientsimilarityjoinmethod.Keywordssimilarityjoin;high-dimensionalspace;mainmemory;datamining;similaritysearch摘要相似连接作为数据挖掘的基元,可被用来大幅度提高相似搜索、数据分
8、析和数据挖掘的速度.大多数研究主要集中在大量基于磁盘数据的高维连接.目前计算机可得到的主存容量越来越大以及对空间连接的有效处理的需求表明,一大类问题的空间连接能够在主存中处理.$-tree是一个新提出的多层索引,已被证实优于其他主存索引.因此,以$-tree为基础,提出了一种空间连接算法$-tree-join,研究*了它的性能,和
