原创力文档- 1、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。。
- 2、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击 这里二次下载。
- 3、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。
基于自监督学习的时序数据异常检测方法研究
Abstract
Timeseriesdatacontainimportantinformationandlaws,whichhelptodetectabnormal
situationsandpreventhazardsintime,andarewidelyusedinvariousfields,e.g.,intrusion
detectioninnetworksecurity,frauddetectioninfinanceandtelecommunication,stockmarket
analysisinthestockmarket,andeventdetectioninthefieldofgeoscience.Duetothefact
thattimeseriesdataislikelytohavenullvaluesorhavemissingdata,itmakesdatalabeling
costly.Furthermore,thecomplexityoftime-seriesdatamakesitmoredifficulttoeffectively
performdatafeatureextraction.Inaddition,themultivariatenatureoftime-seriesdatamakes
ithardtodefineanomalydeterminationrules.Althoughmanyscholarshavealreadystudied
theanomalydetectionmethodsfortimeseriesdata,andachievedalotofresearchresults,
mostofthedetectionmethodsarenoteffectiveinpracticalapplications.Therefore,toaddress
theaboveissues,inthispaper,astudyiscarriedoutonanomalydetectionoftimeseriesdata
usingself-supervisedlearningmethods,andthemainworkisasfollows:
(1)Aimingattheproblemssuchasinsufficienteffectivefeatureextractionin
one-dimensionaltime-seriesdataanomalydetectionmethods,theTSDformermethodbased
onTransformerandtime-seriesdecompositionisproposed.Inthispaper,thecharacterization
informationinthetimeseriesiscapturedbymeansoftimeseriesdecomposition;According
tothecharacteristicsoftimeseriesdata,thefrequencydomaincorrelationisobtainedby
usingthefrequency-domainaugmentedattentionmechanismtoimprovetheprediction
accuracy;Usingreconstructionerrorforanomalydetermination.Multiplesetsofexperiments
areconductedonpublicdatasetsfortheTSDformermethodinthispaper,andtheanomaly
detectionaccuracyreached82.2%.
(2)AnomalyCPCmethodwaspropos
文档评论（0）