ABSTRACT
havemoreorlessdisadvantages.Becausethesimilarityamongsequentialpatternscomesdowntotwomodelsofdifferentlength,byusingthemethodofmeasuringdifferentdimensionsdistance,theauthorputsforwarddynamictimewarpingdistancemeansofsequentialpattern.
ofTemporalAssociationRules(4)Acquirement
fromThethirdstepofminingtemporalassociationrulesishowtogetfrequentpatternssequentialpatters,andthentobuildstrengthenedtemporal
orassociationrules.Incommontemporalassociationrules,the
not,anditsfrequencydependsonobjectsmayappearobjectsandtheappearingtimesofincidents.Becauseofthedifferenceoftimeseriespattern,thefrequencycannotbedecidedbysinglemodel’Sappearingtimes,butbytheamountofsimilarpatterns.Duringtheprocessofcreatingtemporalassociationrules,accordingtotheparticularityoftimeseriespatterns,theauthoroffersthelayeredmeansofgettingtemporalassociationrulesandprovesit.
(5)SimilarityofTimeSeries
Thedissertationclarifies
one
onsimilarityoftimeseriesfromtwoaspects.Onthehand,thedissertationstudiedthesimilarityofone—varietytimeseries.Basedonthesummaryofexistingresearchtimeseries,theauthorputsforwardthe
oftimeseriesandgraphicsimilaritymeasure to -measuresimilarityanalyzesthe
method.Ontheotherhand,thedissertationresearchessimilarityofmultivariatetimeseries.Firstlytheauthoranalyzesthenecessityofresearchingit,andthenthedifficultyinit,finallycomesupwith
basedontwowaystomeasuresimilarityoftimeseries,matrixandsynthesisattribution.
FlatofTemporalassociationrulesoftimeseries(6)Mining
TheminingflatoftemporalassociationrulesoftimeseriesusesJAVAasexploitinglanguages,andhas6modules.Ithasseveralfunctions,suchasloadingdata,timeseriesdatapre—processing,timeseriesdatacompression,timeseriesdatasimilaritymeasure,therequirementoftemporalassociationrulesandtheinterpretationandevaluationoftemporalassociationrules,etc.Thedissertationproveseveryimprovementbyempiricalanalysis,andalsorealizestominetemporalassociationrulesfrom
Combiningwithtimeofseries.theoriesmining
ontemporalassociationrules,thedissertationcarriesoutsystemicresearcheverystep,fromthefirststep,time3