時間序列挖掘中索引與查詢技術(shù)的研究.pdf

1、癤景=墨!I⑩天洋大謦博士學(xué)位論文一級學(xué)科：鯉型差當(dāng)王捏學(xué)科專業(yè)：笪型睦堂皇王猩作者姓名：些壹整——指導(dǎo)教師：立塑盥—監(jiān)一2006年6月ABSTRACTIndexingandqueryingarefundamentalproblemsindataminingThisdissertationaddressestheproblemofindexingandqueryingfortimeseries，anddiscussestherelate

2、dproblemsofthequerymethods，representationandsimilaritymeasuresoftimeseriesBymakingUseoftheproximityquerymethodincomputationalgeometry，thewholematchingquery，patternquery，inversequeryandoutlierdetectionintimeseriesarestudi

3、edThemainresultsofthisdissertationareasfollows：1QuerymethodsoftimeseriesAccordingtothecomponentsoftimeseries，threecategoriesofqueryingareproposed，includingpointquerying，patternquerying，andtimeseriesqueryingAccordingtopro

4、ximitydefinitionincomputationalgeometry，thequeryingoftimeserieshasbeenexpandedtangequery，proximityquery，andclosestpairqueryThedevelopmentoftimeseriesqueryingcanprovideeffectivetechniquesfordatamining2Representationandind

5、exingoftimeseriesBasedontheimportantpointsegmentingoftimeseries，anewsimilaritymeasurebyslopeandlengtharedevelopedFortheindexingoftimeseries，theVoronoidiagramhasbeenusedfortheorganizationoftimeseries，andanewindexingstruct

6、urenamedIC—indexingisproposedfortheinversequeryoftimeseries3SimilarityqueryingintimeseriesWiththeKLsimilaritymeasure，anewalgorithmforwholematchingoftimeserieshasbeendevelopedThepatternqueryingproblemissolvedinoptimaltime

7、basedontheVoronoidiagram4OutlierdetectionintimeseriesBymakinguseoftheVoronoidiagram，anewdensity。basedoutlierdetectionalgorithmisproposed，whichrunsincomputationasO(nlogn)，andhasbeenusedinthelinerpatternoutlierdetectionint