measunngseriespattern’Ssimilaritycannotmeasuretheseriespattern,Ssimilaritvoftwodifferentlengths.
Temporalassociationrulesoftimeseriesarepracticalvaluable,buttheexistingmmmgmethodshavesomeflaws.So,thedissertationfbcusesont11eImprovementandpeffectionoftheminingmethodoftemporalassociationruleSoftimeseries,offeringthetheoreticalmodelsandempiricalanalysis,inordert0gainmorereliabletemporalassociationrulesfromtimeseriesandhelDdecision—making.
Thedissertationaddressestheminingoftemporalassociationrules.Aimingatthefaultinessofeverystep,theauthorsummarizestheexistingrelativeresearch.
offerssolutionsandcarriesoutempiricalanalysis.Thedissertationcanbe
into8chapters,themaincontentareasfollowing.
(1)TimeSeriesDataPre-processing
Timeseriesdatapre—processingisthefirststep0fminingtemporal
thatishowtocleanthenoisedataintimeseries.Inthispart.the
firstdefinesthenoisedata,andthensumsuptheexistingrecognition
ofoutlieroftimeseries,aswellasanalyzestheiradvanta2esand
lastcomesupwiththerecognitionmethodofoutlieroftime
basedonrelativevariancerateoftimeseries.
(2)TimeSeriesDataCompression
Timeseriesdatacompressionisthesecondstepofminingtemporal
rules,whichmeanshowtotransformtimeseriesintoseouential
theauthoranalyzesthenecessity,objectiveandmeaningof
datainminingtemporalassociationrules.Andthenanalyzesthe
compressingways,andthenoffersestimatingsystemtovaluetimeseries
compressmn Aftercomparativeanalysis,choosestimeseriesdata
method,whichisinfavorofmining,andfinallyimprovesthe
ofdivisionpoint.
(3)TimeSeriesDataSimilarityMeasure
SimilaritymeasureofsequentialpatternsistheimportantcontentoftemporaI
rulesoftimeseries.Onlythesimilarityamongpatternsisproperly
acquirementoffrequentpatternsinsequentialpatternsandtemporal
rulesCallbesuccessfullyaccomplished.TheexistingtWOmethodsthendividedassocla“衄rulesauthormethodsdlsadvantages Atseriesassociationpatterns Firstlycompressmgexlstingdatacompressmnreorganizationassociationmeasured,theassociation
2