欢迎来到天天文库
浏览记录
ID:36631682
大小:2.31 MB
页数:64页
时间:2019-05-13
《数据流中离群和变化发现研究》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、j710‘0656分类号:婴31密级:公珏UDC:——学号:Q2窆12璺东南大掌博士学位论文数据流中离群和变化发现研究研究生姓名:扬宣塞导师姓名:处麦握教授申请学位级别王堂燧±论文提交日期窒QQ曼生!!县至!旦学位授予单位峦直去堂学科专业名称j±簋扭廑围论文答辩日期2Q垒且旦学位授予日期.至Q.生且且答辩委员会主席——评阅人2005年11月21日AbstractDataminingisnon-trivi越proce∞thatextractshidden,u丑】Ⅲown,potentiallyusefulknowledgeandpatternfromhugevolumeofdata.
2、Theseyears,manyscholarshavedonedeepandcomprehensiveresearchonthistopic.Recently,withmoreandmoreapplicationsofcomputertechniquesinmanyfields,thereoCcurSanewdatamodel,namelydatastream,whichchaUengescurrentdataminingalgorithms.Themodelofdatastreamexistsinmany鑫el蕊such龉矗珏an∞applications,networkmoni
3、toring,communicationdatamanagement,Webap斑ication,sensornetwork,etc,Begausedatastreamhasitsownclmr毡ctars。such8spotentiallyinfinite.arrivingonlineandrapidly,traditionalminingalgorithmsc8丑notmeetrequestsofthisnewmodel.Dataminingondatast#eamisbecomingahotresearchissue.0utlierdetectionandchangedete
4、ctionarebothimportantdataminingmissionsbe-eanseoftheirfunctionsinknowledgediscovery.ButcurrentalgorithmsforoutHerandchangedetectionaredesignedfortraditionalstaticdatasets。andc∞notbeappliedfordatastream。Tosolvetheproblemof赫馥timeandspacecomplexityofcurrentoutlierdetectionmethod,thisthesispropose
5、safastoutlierdetectionalgorithmfordatastreamsbasedondynamicgrids.whichtakesTavnstfledatastream瞄researchobject.Basedondynamicallygridpartitioningdataspace,thismethodcanfilteroutmostofnormaldataactordingtothedjstributionofdataandapproximationcomputation.whichreducesthesizeofcandidateoutliersandi
6、mprovestheefficiencyofdetectiongreatly.Resultsofexperimentsonsyntheticandrealdatasetsshowpromisingavailabilitiesofoarapproaches.锄1isthesisalsodiscussesdetailsof掣避partitioning,storing,filteringandoutliercomputatiomTimeandspacecomplexityofthisalgorithmarean越yzed,Resultsofexperimentsonsynthetican
7、drealdatasetsshowpromisingavailabflitiesofourapproaches.Therearemanydistributeddatastreamsdatainallkindsofapplications.Thisthesisextendstheproblemofoutlierdetectionindatastreamtothedistributedenvironmentandintroducesanalgorithmforoutlie
此文档下载收益归作者所有