Monitoring Median Queries over Moving Objects.pdfVIP

Vo1．18 No．4 JournalofSouthwestJiaotongUniversity(EnglishEdition) Oct．2010 ArticleID：1005-2429(2010)04-0326-07 DOI：10．3969／j．issn．1005-2429．2010．04．010 MonitoringMedianQueriesoverMovingObjects xuHu(许 浒) ， LUYan—sheng(卢炎生)，LIZhi—cheng(李支成) SchoolofComputerScienceandTechnology，HuazhongUniversityofScienceandTechnology，Wuhan430074，China Abstract Thek-median problem hasattracted anumberofresearchers．However，few ofthem haveconsidered boht hte dynamic environmentnadhteissueofaccuracy．Inthispaper，anewtypeofqueryisstudied，calledcontinuousmedianmonitoring(CMM) query．Itconsidersthe k-medina problem underdynamicenvironmentwith an accuracy guarantee．A continuousgroup nearest neighborbasde (CGB)algorithm nadna averagedistnacemedoid(ADM)algorithm raeproposedtosolvehteCMM problem． ADM isa hillclimbing schemed algorithm and achievesa rapid converging spe ed by checking only qualified candidates． ExperimentsshowhtatADM ismoreefficientthna CGBnadoutperformstheclassicalPAM (partitioningaroundmedoids)nad CLARANS(clusteringlargeapplicationsbasedonrnadomizedserach)algorithmswihtvariouspraametersettings． Keywords Spatialdatabases；Queryprocessing；Neraestneighborquery；k-Medianproblem CLARANScna stillbeinefficientofrevenlragerdata Introduction set．especiallydiskresidentdaat．Estereta1．[ ]de— Thek—median problem isaclassicalproblem in visedamethodcalledFOR (focusonrepresenattives) computationalgeometry．Itretumskcentersfrom aset whichutilizesthehierrachical stmctureofR—tree[引 ． ofpointssuch thatthetot