欢迎来到天天文库
浏览记录
ID:33013161
大小:3.32 MB
页数:65页
时间:2019-02-19
《感知重要点的期货时间序列模式识别的应用分析》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、合肥工业大学硕士学位论文PerceptuallyImportantPoints-BasedFuturesTimeSeriesPatternRecognitionandItsApplicationResearchABSTRACTTimeseriesisanimportantclassoftemporaldataobjects,anditiseasytogetfromthetechnologyandfinancialapplication.Everydataofatimeseriesisobservedinchronologicalor
2、der,thenthetimeseriesisobtainedbyorderedthesedata.Thenatureoftimeseriesincludes:theamountofdataisverylarge,highdimensionsandmustbefrequentlyupdated.Inordertoeffectivelyminingtheimplicitknowledgeofthehistoricaldata,academicandindustrialcirclesuseavarietyoftechniquestore
3、ducethetimesequencecomplexity,andacommonlyusedmethodistoconvertthetimesequenceintoanotherkindofeasytounderstandtheimportantpointsmethodbasedonvisualexpression.Meanwhile,inordertoovercomethedimensiondisasteroftimeseriesdata,acorrespondingindexstructurethatbasedontheexpr
4、essingmethodisproposed.Aimedatidentifyingafuturestimeseriesdatasetwithcertaincharacteristicsandmoreeffectiveidentificationforsomekindofmodel,thisdissertationstudiesitsfeatureidentificationandindexingmethods.Themainresearchworkandachievementsareasfollows.(1)ProposetheZP
5、IP(ZigzagbasedPerceptuallyImportantPoints)importancepointrecognitionmethodbasedonZigzagperceptuallyimportantpoints.ThemethodisbasedonthePIP(PerceptuallyImportantPoints)method.Thealgorithmisoptimizedbythemaximumlocalmethodandstatebalancemethodtograduallyoptimizethelocal
6、lyimportantpoints,andmakeitmoreeffectiveidentificationfuturestimeseriesfeaturepoints.(2)PresenttheFIS(FeatureIndexStructure)featureindexmechanismbasedonthebinarysearchtreetosolvetheinefficientissueofrecognitionprocessinreal-time.Themethodisbasedonabinarysearchtree,inor
7、dertomoresuitableforfuturestimeseriesoffeaturerecognition,weimprovedalgorithmsbyintroducingtheconceptofimportantpointandtheprincipleoftheimportantpointsthathavelargerindexwillmorenearertherootnodeandgetagoodperformance.(3)PutforwardtheFZPIP(FeatureZigzagbasedPerceptual
8、lyImportantPoints)featurerecognitionmethodbasedonthefeatureZigzagperceptuallyimportantpointsbyintroducingthedefinitio
此文档下载收益归作者所有