不确定数据流kTop—k异常点查询算法.pdfVIP

ISSN 1673．94l8 C0DEN JKYTA8 E—mail：fcst@vip．163．tom JournalofFrontiersofComputerScienceandTechnology 1673--9418／2015／09(02)--0172--10 http：／／www．ceaj．org Tel：+86．10 doi：1O．3778~．issn．1673—9418．1405050 不确定数据流kTop—k异常点查询算法 曹科研 ，王国仁 ，韩东红，李硕儒 东北大学 信息科学与工程学院，沈阳 110819 Top一 OutlierDetectionAlgorithm onUncertainDataStream CAOKeyan，WANG Guoren，HAN Donghong，LIShuoru CollegeofInformationScienceandEngineering，NortheasternUniversity,Shenyang110819，China +Correspondingauthor：E—mail：caokeyan@gmail．tom CAO Keyan，WANG Guoren，HAN Donghong,eta1．Top-koutlierdetectionalgorithm onuncertaindatastream． JournalofFrontiersofComputerScienceandTechnology,2015，9(2)：172-181． Abstract：Inrecentyears，alongwithhteappearanceofuncertaindata，outlierdetectiononuncertaindaatstream becomes anew hotspot．However,threeparametersarecontainedintheexistingdefinitionofoutlieronuncertaindata ． itiS verydifficultforuserstosettheseparameters，theusercannotgetthesuitableoutlier ． Mostofthetime，theusers wouldliketogettheobjectswhicharemostlikelytobeoutliers．Thispaperproposesthetop-koutlierdetectionon uncertaindatastream．Theproposedmethodpmnesobjectsbasedontheestimationoftherangeofprobabilitiesbeing outlierandreducessomeunnecessary computation．Meanwhile，thispaperproposestheincrementalmethod for computingtherangeofprobabilitiestoimproveeffi ciency．Finally,theperfomr anceoftheproposedmethodisveli- fledthroughanumberofsimulationexperimentsonrealandsyntheticdatasets． Keywords：uncertaindata；datamining；outlier；top—k 摘 要：近几年，随着数据流和不确定数据的产生，不确定数据流上的异常点检测成为新的研究热点。然而， 现有的不确定数据的异常点定