Article(id=1156908300735439743, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1156908295593223005, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2308252, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1697990400000, receivedDateStr=2023-10-23, revisedDate=1728316800000, revisedDateStr=2024-10-08, acceptedDate=null, acceptedDateStr=null, onlineDate=1753758033212, onlineDateStr=2025-07-29, pubDate=1736265600000, pubDateStr=2025-01-08, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1753758033212, onlineIssueDateStr=2025-07-29, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1753758033212, creator=13701087609, updateTime=1753758033212, updator=13701087609, issue=Issue{id=1156908295593223005, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='1', pageStart='1', pageEnd='438', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=0, createTime=1753758031985, creator=13701087609, updateTime=1765425680602, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1205845960933049001, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1156908295593223005, language=EN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1205845960933049002, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1156908295593223005, language=CN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=252, endPage=261, ext={EN=ArticleExt(id=1156908302018896771, articleId=1156908300735439743, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=Multivariate Time Series Classification Method Based on Shapelets, columnId=1156262729162810294, journalTitle=Science Technology and Engineering, columnName=Papers·Automation and Computational Technology, runingTitle=null, highlight=null, articleAbstract=

Multivariate time series classification is a key problem in many fields, but the current research on multivariate time series classification is faced with some problems, such as high dimensionality of original data, low accuracy, and lack of interpretability, which limits the performance improvement of models and makes it difficult to meet the actual requirements. Aiming at above problem, a multivariate time series classification method based on Shapelets was proposed. Firstly, unsupervised Shapelet learning of adaptive neighbors was used to automatically learn significant multivariate Shapelets by combining Shapelets transform and adaptive weights. Then, the method was combined with Shapelet similarity and class label constraint to enhance the interpretability and classification accuracy of the model. Finally, the optimization strategy of the model was proposed to obtain the best Shapelets to further improve the classification accuracy of the model. Three different types of 11 algorithms were compared on 11 public data sets, and the experimental results show that the proposed algorithm has high classification accuracy.

, correspAuthors=null, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, authorCompany=null, fund=null, authors=null, authorsList=Wei-na WANG, Ming-li LI), CN=ArticleExt(id=1156908364044263521, articleId=1156908300735439743, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=基于Shapelets的多元时间序列分类方法, columnId=1156262729783567290, journalTitle=科学技术与工程, columnName=论文·自动化技术、计算机技术, runingTitle=null, highlight=null, articleAbstract=

多元时间序列分类是众多领域的关键问题,但是当前多元时序分类研究面临着原始数据高维、精度不足、可解释性缺乏等问题,这使得模型性能提升受限,准确率难以满足实际需求。针对上述问题,提出基于Shapelets的多元时间序列分类方法。首先,利用自适应邻居的无监督Shapelet学习将Shapelet变换与自适应权重结合,用于自动学习显著多元Shapelets;然后,将该方法与Shapelet相似性和类标约束项结合,增强模型可解释性和分类准确性;最后,提出模型的优化策略,用以获取最优的Shapelets,进一步提高模型的分类精度。与3种不同类型11个算法在11个公开数据集上进行比较,实验结果表明提出算法具有较高的分类精度。

, correspAuthors=null, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=ZOQCxJNhhubhilrvUHEeSA==, magXml=8anooC0b/LEnvYAFIAr/cA==, pdfUrl=null, pdf=atOm0g0A4POOonLYGimBDw==, pdfFileSize=4571953, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=Bv0GxcvilT8MBNC33xIQSw==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=YNZlLKhyyJ5T0tHEhZ0hkw==, mapNumber=null, authorCompany=null, fund=null, authors=

王威娜(1981—),女,汉族,吉林吉林人,博士,教授。研究方向:时间序列分析、数据挖掘。E-mail:

, authorsList=王威娜, 李明莉)}, authors=[Author(id=1205916736478572643, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=wangweina@jlict.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1205916736545681510, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, authorId=1205916736478572643, language=EN, stringName=Wei-na WANG, firstName=Wei-na, middleName=null, lastName=WANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Information and Control Engineering, Jilin Institute of Chemical Technology, Jilin 132022, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1205916736616984679, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, authorId=1205916736478572643, language=CN, stringName=王威娜, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=吉林化工学院信息与控制工程学院, 吉林 132022, bio={"content":"

王威娜(1981—),女,汉族,吉林吉林人,博士,教授。研究方向:时间序列分析、数据挖掘。E-mail:

"}, bioImg=null, bioContent=

王威娜(1981—),女,汉族,吉林吉林人,博士,教授。研究方向:时间序列分析、数据挖掘。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1205916736348549215, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, xref=null, ext=[AuthorCompanyExt(id=1205916736361132128, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, companyId=1205916736348549215, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Information and Control Engineering, Jilin Institute of Chemical Technology, Jilin 132022, China), AuthorCompanyExt(id=1205916736373715041, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, companyId=1205916736348549215, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=吉林化工学院信息与控制工程学院, 吉林 132022)])]), Author(id=1205916736700870761, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1205916736763785323, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, authorId=1205916736700870761, language=EN, stringName=Ming-li LI, firstName=Ming-li, middleName=null, lastName=LI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Information and Control Engineering, Jilin Institute of Chemical Technology, Jilin 132022, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1205916736826699884, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, authorId=1205916736700870761, language=CN, stringName=李明莉, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=吉林化工学院信息与控制工程学院, 吉林 132022, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1205916736348549215, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, xref=null, ext=[AuthorCompanyExt(id=1205916736361132128, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, companyId=1205916736348549215, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Information and Control Engineering, Jilin Institute of Chemical Technology, Jilin 132022, China), AuthorCompanyExt(id=1205916736373715041, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, companyId=1205916736348549215, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=吉林化工学院信息与控制工程学院, 吉林 132022)])])], keywords=[Keyword(id=1205916736948334701, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, orderNo=1, keyword=multivariate time series), Keyword(id=1205916737032220782, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, orderNo=2, keyword=multivariate time series classification), Keyword(id=1205916737124495471, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, orderNo=3, keyword=Shapelets learning), Keyword(id=1205916737191604336, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, orderNo=4, keyword=optimization strategy), Keyword(id=1205916738361815153, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, orderNo=1, keyword=多元时间序列), Keyword(id=1205916738433118322, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, orderNo=2, keyword=多元时间序列分类), Keyword(id=1205916738504421491, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, orderNo=3, keyword=Shapelets学习), Keyword(id=1205916738571530356, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, orderNo=4, keyword=优化策略)], refs=[Reference(id=1205916740299583624, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2022, volume=36, issue=3, pageStart=1015, pageEnd=1061, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=Feremans L, Cule B, Goethals B, journalName=Data Mining and Knowledge Discovery, refType=null, unstructuredReference=Feremans L, Cule B, Goethals B. PETSC: pattern-based embedding for time series classification[J]. Data Mining and Knowledge Discovery, 2022, 36(3): 1015-1061., articleTitle=PETSC: pattern-based embedding for time series classification, refAbstract=null), Reference(id=1205916740366692489, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=49, issue=4, pageStart=790, pageEnd=804, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=张旭, 张亮, 金博, journalName=自动化学报, refType=null, unstructuredReference=张旭, 张亮, 金博, 等. 基于不确定性的多元时间序列分类算法研究[J]. 自动化学报, 2023, 49(4): 790-804., articleTitle=基于不确定性的多元时间序列分类算法研究, refAbstract=null), Reference(id=1205916740421218442, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=49, issue=4, pageStart=790, pageEnd=804, url=null, language=null, rfNumber=[2], rfOrder=2, authorNames=Zhang Xu, Zhang Liang, Jin Bo, journalName=Acta Automatica Sinica, refType=null, unstructuredReference=Zhang Xu, Zhang Liang, Jin Bo, et al. Research on classification algorithm of multivariate time series based on uncertainty[J]. Acta Automatica Sinica, 2023, 49(4): 790-804., articleTitle=Research on classification algorithm of multivariate time series based on uncertainty, refAbstract=null), Reference(id=1205916740488327307, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=22, issue=1, pageStart=1, pageEnd=24, url=null, language=null, rfNumber=[3], rfOrder=3, authorNames=Velasco I, Sipols A, De Blas C S, journalName=BioMedical Engineering OnLine, refType=null, unstructuredReference=Velasco I, Sipols A, De Blas C S, et al. Motor imagery EEG signal classification with a multivariate time series approach[J]. BioMedical Engineering OnLine, 2023, 22(1): 1-24., articleTitle=Motor imagery EEG signal classification with a multivariate time series approach, refAbstract=null), Reference(id=1205916740584796300, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=null, pageStart=2301, pageEnd=null, url=null, language=null, rfNumber=[4], rfOrder=4, authorNames=Lee J, Youn H L, Poon J, journalName=ArXiv Preprint ArXiv, refType=null, unstructuredReference=Lee J, Youn H L, Poon J, et al. StockEmotions: discover investor emotions for financial sentiment analysis and multivariate time series[J]. ArXiv Preprint ArXiv, 2023: 2301.09279., articleTitle=StockEmotions: discover investor emotions for financial sentiment analysis and multivariate time series, refAbstract=null), Reference(id=1205916740647710861, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2017, volume=477, issue=null, pageStart=149, pageEnd=160, url=null, language=null, rfNumber=[5], rfOrder=5, authorNames=Yan Y, Zhang S, Tang J, journalName=Physica A: Statistical Mechanics and its Applications, refType=null, unstructuredReference=Yan Y, Zhang S, Tang J, et al. Understanding characteristics in multivariate traffic flow time series from complex network structure[J]. Physica A: Statistical Mechanics and its Applications, 2017, 477: 149-160., articleTitle=Understanding characteristics in multivariate traffic flow time series from complex network structure, refAbstract=null), Reference(id=1205916740710625422, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=7, issue=5, pageStart=164, pageEnd=null, url=null, language=null, rfNumber=[6], rfOrder=6, authorNames=Rahman M M, Farahani M A, Wuest T, journalName=Journal of Manufacturing and Materials Processing, refType=null, unstructuredReference=Rahman M M, Farahani M A, Wuest T. Multivariate time-series classification of critical events from industrial drying hopper operations: a deep learning approach[J]. Journal of Manufacturing and Materials Processing, 2023, 7(5): 164., articleTitle=Multivariate time-series classification of critical events from industrial drying hopper operations: a deep learning approach, refAbstract=null), Reference(id=1205916740769345679, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2020, volume=20, issue=5, pageStart=1747, pageEnd=1760, url=null, language=null, rfNumber=[7], rfOrder=7, authorNames=Gupta A, Gupta H P, Biswas B, journalName=IEEE Transactions on Mobile Computing, refType=null, unstructuredReference=Gupta A, Gupta H P, Biswas B, et al. A fault-tolerant early classification approach for human activities using multivariate time series[J]. IEEE Transactions on Mobile Computing, 2020, 20(5): 1747-1760., articleTitle=A fault-tolerant early classification approach for human activities using multivariate time series, refAbstract=null), Reference(id=1205916740836454544, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2022, volume=45, issue=4, pageStart=4981, pageEnd=4996, url=null, language=null, rfNumber=[8], rfOrder=8, authorNames=Zhang N, Sun S, journalName=IEEE Transactions on Pattern Analysis and Machine Intelligence, refType=null, unstructuredReference=Zhang N, Sun S. Multiview unsupervised shapelet learning for multivariate time series clustering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 45(4): 4981-4996., articleTitle=Multiview unsupervised shapelet learning for multivariate time series clustering, refAbstract=null), Reference(id=1205916740911952017, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2010, volume=43, issue=11, pageStart=3787, pageEnd=3794, url=null, language=null, rfNumber=[9], rfOrder=9, authorNames=Orsenigo C, Vercellis C, journalName=Pattern Recognition, refType=null, unstructuredReference=Orsenigo C, Vercellis C. Combining discrete SVM and fixed cardinality warping distances for multivariate time series classification[J]. Pattern Recognition, 2010, 43(11): 3787-3794., articleTitle=Combining discrete SVM and fixed cardinality warping distances for multivariate time series classification, refAbstract=null), Reference(id=1205916740979060882, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=3617, pageEnd=3624, url=null, language=null, rfNumber=[10], rfOrder=10, authorNames=Cheng Z, Yang Y, Wang W, journalName=Proceedings of the AAAI Conference on Artificial Intelligence, refType=null, unstructuredReference=Cheng Z, Yang Y, Wang W, et al. Time2graph: revisiting time series modeling with dynamic Shapelets[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2020: 3617-3624., articleTitle=Time2graph: revisiting time series modeling with dynamic Shapelets, refAbstract=null), Reference(id=1205916741037781139, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=158, pageEnd=166, url=null, language=null, rfNumber=[11], rfOrder=11, authorNames=Cho S, Chang W, Lee G, journalName=Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, refType=null, unstructuredReference=Cho S, Chang W, Lee G, et al. Interpreting internal activation patterns in deep temporal neural networks by finding prototypes[C]// Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. New York: Association for Computing Machinery, 2021: 158-166., articleTitle=Interpreting internal activation patterns in deep temporal neural networks by finding prototypes, refAbstract=null), Reference(id=1205916741092307092, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2019, volume=7, issue=null, pageStart=139580, pageEnd=139591, url=null, language=null, rfNumber=[12], rfOrder=12, authorNames=Wu G, Zhang H, He Y, journalName=IEEE Access, refType=null, unstructuredReference=Wu G, Zhang H, He Y, et al. Learning Kullback-Leibler divergence-based Gaussian model for multivariate time series classification[J]. IEEE Access, 2019, 7: 139580-139591., articleTitle=Learning Kullback-Leibler divergence-based Gaussian model for multivariate time series classification, refAbstract=null), Reference(id=1205916741155221653, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=1994, volume=null, issue=null, pageStart=359, pageEnd=370, url=null, language=null, rfNumber=[13], rfOrder=13, authorNames=Berndt D J, Clifford J, journalName=Proceedings of the 3rd International Conference on Knowledge Discovery and Data Mining, refType=null, unstructuredReference=Berndt D J, Clifford J. Using dynamic time warping to find patterns in time series[C]// Proceedings of the 3rd International Conference on Knowledge Discovery and Data Mining. Palo Alto: AAAI, 1994: 359-370., articleTitle=Using dynamic time warping to find patterns in time series, refAbstract=null), Reference(id=1205916741222330518, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=30, issue=null, pageStart=369, pageEnd=373, url=null, language=null, rfNumber=[14], rfOrder=14, authorNames=Zhao B, Xing H, Wang X, journalName=IEEE Signal Processing Letters, refType=null, unstructuredReference=Zhao B, Xing H, Wang X, et al. Classification-oriented distributed semantic communication for multivariate time series[J]. IEEE Signal Processing Letters, 2023, 30: 369-373., articleTitle=Classification-oriented distributed semantic communication for multivariate time series, refAbstract=null), Reference(id=1205916741289439383, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=53, pageEnd=69, url=null, language=null, rfNumber=[15], rfOrder=15, authorNames=Le X M, Tran M T, Huynh V N, journalName=Joint European Conference on Machine Learning and Knowledge Discovery in Databases, refType=null, unstructuredReference=Le X M, Tran M T, Huynh V N. Learning perceptual position-aware shapelets for time series classification[C]// Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Cham: Springer Nature Switzerland, 2022: 53-69., articleTitle=Learning perceptual position-aware shapelets for time series classification, refAbstract=null), Reference(id=1205916741360742552, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2022, volume=33, issue=24, pageStart=2990, pageEnd=2996, url=null, language=null, rfNumber=[16], rfOrder=16, authorNames=李可, 燕晗, 顾杰斐, journalName=中国机械工程, refType=null, unstructuredReference=李可, 燕晗, 顾杰斐, 等. 基于Shapelets时间序列的多源迁移学习滚动轴承故障诊断方法[J]. 中国机械工程, 2022, 33(24): 2990-2996., articleTitle=基于Shapelets时间序列的多源迁移学习滚动轴承故障诊断方法, refAbstract=null), Reference(id=1205916741444628633, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2022, volume=33, issue=24, pageStart=2990, pageEnd=2996, url=null, language=null, rfNumber=[16], rfOrder=17, authorNames=Li Ke, Yan Han, Gu Jiefei, journalName=China Mechanical Engineering, refType=null, unstructuredReference=Li Ke, Yan Han, Gu Jiefei, et al. Multi-source transfer learning rolling bearing fault diagnosis method based on Shapelets time series[J]. China Mechanical Engineering, 2022, 33(24): 2990-2996., articleTitle=Multi-source transfer learning rolling bearing fault diagnosis method based on Shapelets time series, refAbstract=null), Reference(id=1205916741499154586, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2019, volume=132, issue=null, pageStart=76, pageEnd=86, url=null, language=null, rfNumber=[17], rfOrder=18, authorNames=Li G, Yan W, Wu Z, journalName=Expert Systems with Applications, refType=null, unstructuredReference=Li G, Yan W, Wu Z. Discovering Shapelets with key points in time series classification[J]. Expert Systems with Applications, 2019, 132: 76-86., articleTitle=Discovering Shapelets with key points in time series classification, refAbstract=null), Reference(id=1205916741574652059, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2022, volume=38, issue=6, pageStart=2056, pageEnd=2074, url=null, language=null, rfNumber=[18], rfOrder=19, authorNames=Hssayni E H, Joudar N E, Ettaouil M, journalName=Computational Intelligence, refType=null, unstructuredReference=Hssayni E H, Joudar N E, Ettaouil M. A deep learning framework for time series classification using normal cloud representation and convolutional neural network optimization[J]. Computational Intelligence, 2022, 38(6): 2056-2074., articleTitle=A deep learning framework for time series classification using normal cloud representation and convolutional neural network optimization, refAbstract=null), Reference(id=1205916741658538140, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2018, volume=66, issue=6, pageStart=4788, pageEnd=4797, url=null, language=null, rfNumber=[19], rfOrder=20, authorNames=Liu C L, Hsaio W H, Tu Y C, journalName=IEEE Transactions on Industrial Electronics, refType=null, unstructuredReference=Liu C L, Hsaio W H, Tu Y C. Time series classification with multivariate convolutional neural network[J]. IEEE Transactions on Industrial Electronics, 2018, 66(6): 4788-4797., articleTitle=Time series classification with multivariate convolutional neural network, refAbstract=null), Reference(id=1205916742803583133, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=2103, pageEnd=null, url=null, language=null, rfNumber=[20], rfOrder=21, authorNames=Liu M, Ren S, Ma S, journalName=ArXiv Preprint ArXiv, refType=null, unstructuredReference=Liu M, Ren S, Ma S, et al. Gated transformer networks for multivariate time series classification[J]. ArXiv Preprint ArXiv, 2021: 2103. 14438., articleTitle=Gated transformer networks for multivariate time series classification, refAbstract=null), Reference(id=1205916742895857822, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2021, volume=33, issue=18, pageStart=45, pageEnd=47, url=null, language=null, rfNumber=[21], rfOrder=22, authorNames=孙其法, journalName=信息与电脑(理论版), refType=null, unstructuredReference=孙其法. 基于多样化Shapelets的不平衡时间序列分类方法[J]. 信息与电脑(理论版), 2021, 33(18): 45-47., articleTitle=基于多样化Shapelets的不平衡时间序列分类方法, refAbstract=null), Reference(id=1205916742950383775, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2021, volume=33, issue=18, pageStart=45, pageEnd=47, url=null, language=null, rfNumber=[21], rfOrder=23, authorNames=Sun Qifa, journalName=Information and Computers (Theory), refType=null, unstructuredReference=Sun Qifa. Unbalanced time series classification method based on diversified Shapelets[J]. Information and Computers (Theory), 2021, 33(18): 45-47., articleTitle=Unbalanced time series classification method based on diversified Shapelets, refAbstract=null), Reference(id=1205916743017492640, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2009, volume=null, issue=null, pageStart=947, pageEnd=956, url=null, language=null, rfNumber=[22], rfOrder=24, authorNames=Ye L, Keogh E, journalName=International Conference on Knowledge Discovery and Data Mining, refType=null, unstructuredReference=Ye L, Keogh E. Time series Shapelets: a new primitive for data mining[C]// International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2009: 947-956., articleTitle=Time series Shapelets: a new primitive for data mining, refAbstract=null), Reference(id=1205916743084601505, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=null, pageStart=2302, pageEnd=null, url=null, language=null, rfNumber=[23], rfOrder=25, authorNames=Liang Z, Wang H, journalName=ArXiv Preprint ArXiv, refType=null, unstructuredReference=Liang Z, Wang H. FedST: federated Shapelet transformation for interpretable time series classification[J]. ArXiv Preprint ArXiv, 2023: 2302. 10631., articleTitle=FedST: federated Shapelet transformation for interpretable time series classification, refAbstract=null), Reference(id=1205916743168487586, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=53, issue=5, pageStart=5240, pageEnd=5258, url=null, language=null, rfNumber=[24], rfOrder=26, authorNames=Yang J, Jing S, Huang G, journalName=Applied Intelligence, refType=null, unstructuredReference=Yang J, Jing S, Huang G. Accurate and fast time series classification based on compressed random shapelet forest[J]. Applied Intelligence, 2023, 53(5): 5240-5258., articleTitle=Accurate and fast time series classification based on compressed random shapelet forest, refAbstract=null), Reference(id=1205916743264956579, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2012, volume=null, issue=null, pageStart=289, pageEnd=297, url=null, language=null, rfNumber=[25], rfOrder=27, authorNames=Lines J, Davis L M, Hills J, journalName=Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, refType=null, unstructuredReference=Lines J, Davis L M, Hills J, et al. A Shapelet transform for time series classification[C]// Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: Association for Computing Machinery, 2012: 289-297., articleTitle=A Shapelet transform for time series classification, refAbstract=null), Reference(id=1205916743340454052, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2014, volume=null, issue=null, pageStart=392, pageEnd=401, url=null, language=null, rfNumber=[26], rfOrder=28, authorNames=Grabocka J, Schilling N, Wistuba M, journalName=Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, refType=null, unstructuredReference=Grabocka J, Schilling N, Wistuba M, et al. Learning time-series shapelets[C]// Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: Association for Computing Machinery, 2014: 392-401., articleTitle=Learning time-series shapelets, refAbstract=null), Reference(id=1205916743415951525, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=653, pageEnd=664, url=null, language=null, rfNumber=[27], rfOrder=29, authorNames=Guillaume A, Vrain C, Elloumi W, journalName=International Conference on Pattern Recognition and Artificial Intelligence, refType=null, unstructuredReference=Guillaume A, Vrain C, Elloumi W. Random dilated shapelet transform: a new approach for time series shapelets[C]// International Conference on Pattern Recognition and Artificial Intelligence. Cham: Springer International Publishing, 2022: 653-664., articleTitle=Random dilated shapelet transform: a new approach for time series shapelets, refAbstract=null), Reference(id=1205916743474671782, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=53, issue=14, pageStart=17985, pageEnd=18001, url=null, language=null, rfNumber=[28], rfOrder=30, authorNames=Chen J, Wan Y, journalName=Applied Intelligence, refType=null, unstructuredReference=Chen J, Wan Y. Localized Shapelets selection for interpretable time series classification[J]. Applied Intelligence, 2023, 53(14): 17985-18001., articleTitle=Localized Shapelets selection for interpretable time series classification, refAbstract=null), Reference(id=1205916743541780647, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2022, volume=12, issue=17, pageStart=8685, pageEnd=null, url=null, language=null, rfNumber=[29], rfOrder=31, authorNames=Liu H Y, Gao Z Z, Wang Z H, journalName=Applied Sciences, refType=null, unstructuredReference=Liu H Y, Gao Z Z, Wang Z H, et al. Time series classification with Shapelet and canonical features[J]. Applied Sciences, 2022, 12(17): 8685., articleTitle=Time series classification with Shapelet and canonical features, refAbstract=null), Reference(id=1205916743650832552, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=8, pageStart=3345, pageEnd=3353, url=null, language=null, rfNumber=[30], rfOrder=32, authorNames=王威娜, 胡佳利, 任艳, journalName=科学技术与工程, refType=null, unstructuredReference=王威娜, 胡佳利, 任艳. 基于优化Shapelet的时间序列分类方法[J]. 科学技术与工程, 2023, 23(8): 3345-3353., articleTitle=基于优化Shapelet的时间序列分类方法, refAbstract=null), Reference(id=1205916743734718633, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=8, pageStart=3345, pageEnd=3353, url=null, language=null, rfNumber=[30], rfOrder=33, authorNames=Wang Weina, Hu Jiali, Ren Yan, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Wang Weina, Hu Jiali, Ren Yan. Time series classification method based on optimized Shapelet[J]. Science Technology and Engineering, 2023, 23(8): 3345-3353., articleTitle=Time series classification method based on optimized Shapelet, refAbstract=null), Reference(id=1205916743810216106, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=50, issue=2, pageStart=146, pageEnd=157, url=null, language=null, rfNumber=[31], rfOrder=34, authorNames=李晨, 万源, journalName=计算机科学, refType=null, unstructuredReference=李晨, 万源. 基于优化和两阶段筛选的时间序列Shapelets提取研究[J]. 计算机科学, 2023, 50(2): 146-157., articleTitle=基于优化和两阶段筛选的时间序列Shapelets提取研究, refAbstract=null), Reference(id=1205916743889907883, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=50, issue=2, pageStart=146, pageEnd=157, url=null, language=null, rfNumber=[31], rfOrder=35, authorNames=Li Chen, Wan Yuan, journalName=Computer Science, refType=null, unstructuredReference=Li Chen, Wan Yuan. Research on extraction of time series Shapelets based on optimization and two-stage screening[J]. Computer Science, 2023, 50(2): 146-157., articleTitle=Research on extraction of time series Shapelets based on optimization and two-stage screening, refAbstract=null), Reference(id=1205916743961211052, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=52, issue=6, pageStart=887, pageEnd=896, url=null, language=null, rfNumber=[32], rfOrder=36, authorNames=杨骏, 敬思远, 钟勇, journalName=电子科技大学学报, refType=null, unstructuredReference=杨骏, 敬思远, 钟勇. 面向时间序列有序分类的Shapelet抽取算法[J]. 电子科技大学学报, 2023, 52(6): 887-896., articleTitle=面向时间序列有序分类的Shapelet抽取算法, refAbstract=null), Reference(id=1205916744019931309, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=52, issue=6, pageStart=887, pageEnd=896, url=null, language=null, rfNumber=[32], rfOrder=37, authorNames=Yang Jun, Jing Siyuan, Zhong Yong, journalName=Journal of University of Electronic Science and Technology, refType=null, unstructuredReference=Yang Jun, Jing Siyuan, Zhong Yong. Shapelet extraction algorithm for ordered classification of time series[J]. Journal of University of Electronic Science and Technology, 2023, 52(6): 887-896., articleTitle=Shapelet extraction algorithm for ordered classification of time series, refAbstract=null), Reference(id=1205916744103817390, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2019, volume=46, issue=1, pageStart=29, pageEnd=35, url=null, language=null, rfNumber=[33], rfOrder=38, authorNames=闫汶和, 李桂玲, journalName=计算机科学, refType=null, unstructuredReference=闫汶和, 李桂玲. 基于Shapelet的时间序列分类研究[J]. 计算机科学, 2019, 46(1): 29-35., articleTitle=基于Shapelet的时间序列分类研究, refAbstract=null), Reference(id=1205916744162537647, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2019, volume=46, issue=1, pageStart=29, pageEnd=35, url=null, language=null, rfNumber=[33], rfOrder=39, authorNames=Yan Wenhe, Li Guiling, journalName=Computer Science, refType=null, unstructuredReference=Yan Wenhe, Li Guiling. Research on time series classification based on Shapelet[J]. Computer Science, 2019, 46(1): 29-35., articleTitle=Research on time series classification based on Shapelet, refAbstract=null), Reference(id=1205916744212869296, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2022, volume=45, issue=4, pageStart=4981, pageEnd=4996, url=null, language=null, rfNumber=[34], rfOrder=40, authorNames=Zhang N, Sun S, journalName=IEEE Transactions on Pattern Analysis and Machine Intelligence, refType=null, unstructuredReference=Zhang N, Sun S. Multiview unsupervised Shapelet learning for multivariate time series clustering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 45(4): 4981-4996., articleTitle=Multiview unsupervised Shapelet learning for multivariate time series clustering, refAbstract=null), Reference(id=1205916744284172465, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2014, volume=28, issue=null, pageStart=851, pageEnd=881, url=null, language=null, rfNumber=[35], rfOrder=41, authorNames=Hills J, Lines J, Baranauskas E, journalName=Data Mining and Knowledge Discovery, refType=null, unstructuredReference=Hills J, Lines J, Baranauskas E, et al. Classification of time series by Shapelet transformation[J]. Data Mining and Knowledge Discovery, 2014, 28: 851-881., articleTitle=Classification of time series by Shapelet transformation, refAbstract=null), Reference(id=1205916744338698418, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2015, volume=null, issue=null, pageStart=289, pageEnd=297, url=null, language=null, rfNumber=[36], rfOrder=42, authorNames=Shokoohi-Yekta M, Wang J, Keogh E J, journalName=Proceedings of the 2015 SIAM International Conference on Data Mining, refType=null, unstructuredReference=Shokoohi-Yekta M, Wang J, Keogh E J. On the non-trivial generalization of dynamic time warping to the multi-dimensional case[C]// Proceedings of the 2015 SIAM International Conference on Data Mining. Vancouver, Canada: SIAM, 2015: 289-297., articleTitle=On the non-trivial generalization of dynamic time warping to the multi-dimensional case, refAbstract=null), Reference(id=1205916744405807283, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=8375, pageEnd=8383, url=null, language=null, rfNumber=[37], rfOrder=43, authorNames=Li G, Choi B, Xu J, journalName=Proceedings of the AAAI Conference on Artificial Intelligence, refType=null, unstructuredReference=Li G, Choi B, Xu J, et al. Shapenet: a Shapelet-neural network approach for multivariate time series classification[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2021: 8375-8383., articleTitle=Shapenet: a Shapelet-neural network approach for multivariate time series classification, refAbstract=null), Reference(id=1205916744468721844, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2016, volume=30, issue=5, pageStart=1053, pageEnd=1085, url=null, language=null, rfNumber=[38], rfOrder=44, authorNames=Karlsson I, Papapetrou P, Boström H, journalName=Data Mining and Knowledge Discovery, refType=null, unstructuredReference=Karlsson I, Papapetrou P, Boström H. Generalized random Shapelet forests[J]. Data Mining and Knowledge Discovery, 2016, 30(5): 1053-1085., articleTitle=Generalized random Shapelet forests, refAbstract=null), Reference(id=1205916744535830709, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2021, volume=39, issue=2, pageStart=205, pageEnd=214, url=null, language=null, rfNumber=[39], rfOrder=45, authorNames=宋奎勇, 王念滨, 王红滨, journalName=吉林大学学报 (信息科学版), refType=null, unstructuredReference=宋奎勇, 王念滨, 王红滨. 基于 Shapelets 的多变量 DS 证据加权集成分类[J]. 吉林大学学报 (信息科学版), 2021, 39(2): 205-214., articleTitle=基于 Shapelets 的多变量 DS 证据加权集成分类, refAbstract=null), Reference(id=1205916744649076918, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2021, volume=39, issue=2, pageStart=205, pageEnd=214, url=null, language=null, rfNumber=[39], rfOrder=46, authorNames=Song Kuiyong, Wang Nianbin, Wang Hongbin, journalName=Journal of Jilin University (Information Science Edition), refType=null, unstructuredReference=Song Kuiyong, Wang Nianbin, Wang Hongbin. Multivariate DS evidence weighted ensemble classification based on Shapelets[J]. Journal of Jilin University (Information Science Edition), 2021, 39(2): 205-214., articleTitle=Multivariate DS evidence weighted ensemble classification based on Shapelets, refAbstract=null), Reference(id=1205916744737157303, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=38, issue=2, pageStart=386, pageEnd=400, url=null, language=null, rfNumber=[40], rfOrder=47, authorNames=詹熙, 黎维, 潘志松, journalName=数据采集与处理, refType=null, unstructuredReference=詹熙, 黎维, 潘志松. Multi-Shapelet: 一种基于Shapelet的多变量时间序列分类方法[J]. 数据采集与处理, 2023, 38(2): 386-400., articleTitle=Multi-Shapelet: 一种基于Shapelet的多变量时间序列分类方法, refAbstract=null), Reference(id=1205916744816849080, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=38, issue=2, pageStart=386, pageEnd=400, url=null, language=null, rfNumber=[40], rfOrder=48, authorNames=Zhan Xi, Li Wei, Pan Zhisong, journalName=Data Acquisition and Processing, refType=null, unstructuredReference=Zhan Xi, Li Wei, Pan Zhisong. Multi-Shapelet: a multivariable time series classification method based on Shapelet[J]. Data Acquisition and Processing, 2023, 38(2): 386-400., articleTitle=Multi-Shapelet: a multivariable time series classification method based on Shapelet, refAbstract=null), Reference(id=1205916744930095289, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=1811, pageEnd=null, url=null, language=null, rfNumber=[41], rfOrder=49, authorNames=Bagnall A, Dau H A, Lines J, journalName=ArXiv Preprint ArXiv, refType=null, unstructuredReference=Bagnall A, Dau H A, Lines J, et al. The UEA multivariate time series classification archive, 2018[J]. ArXiv Preprint ArXiv, 2018: 1811.00075., articleTitle=The UEA multivariate time series classification archive, 2018, refAbstract=null), Reference(id=1205916744993009850, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=6845, pageEnd=6852, url=null, language=null, rfNumber=[42], rfOrder=50, authorNames=Zhang X, Gao Y, Lin J, journalName=Proceedings of the AAAI Conference on Artificial Intelligence, refType=null, unstructuredReference=Zhang X, Gao Y, Lin J, et al. Tapnet: multivariate time series classification with attentional prototypical network[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2020: 6845-6852., articleTitle=Tapnet: multivariate time series classification with attentional prototypical network, refAbstract=null), Reference(id=1205916745110450363, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=1565, pageEnd=1570, url=null, language=null, rfNumber=[43], rfOrder=51, authorNames=Zuo J, Zeitouni K, Taher Y, journalName=null, refType=null, unstructuredReference=Zuo J, Zeitouni K, Taher Y. Smate:semi-supervised spatio-temporal representation learning on multivariate time series[C]//2021 IEEE International Conference on Data Mining (ICDM). Piscataway: IEEE, 2021: 1565-1570., articleTitle=Smate:semi-supervised spatio-temporal representation learning on multivariate time series[C], refAbstract=null), Reference(id=1205916745185947836, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2022, volume=610, issue=null, pageStart=472, pageEnd=487, url=null, language=null, rfNumber=[44], rfOrder=52, authorNames=Chen R, Yan X, Wang S, journalName=Information Sciences, refType=null, unstructuredReference=Chen R, Yan X, Wang S, et al. DA-Net: dual-attention network for multivariate time series classification[J]. Information Sciences, 2022, 610: 472-487., articleTitle=DA-Net: dual-attention network for multivariate time series classification, refAbstract=null), Reference(id=1205916745253056701, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, doi=null, pmid=null, pmcid=null, year=2023, volume=639, issue=null, pageStart=119009, pageEnd=null, url=null, language=null, rfNumber=[45], rfOrder=53, authorNames=Du M, Wei Y, Zheng X, journalName=Information Sciences, refType=null, unstructuredReference=Du M, Wei Y, Zheng X, et al. Multi-feature based network for multivariate time series classification[J]. Information Sciences, 2023, 639: 119009., articleTitle=Multi-feature based network for multivariate time series classification, refAbstract=null)], funds=[Fund(id=1205916740022759557, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, awardId=62266046, language=CN, fundingSource=国家自然科学基金(62266046), fundOrder=null, country=null), Fund(id=1205916740098257030, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, awardId=YDZJ202201ZYTS603, language=CN, fundingSource=吉林省自然科学基金(YDZJ202201ZYTS603), fundOrder=null, country=null), Fund(id=1205916740148588679, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, awardId=JJKH20230281KJ, language=CN, fundingSource=吉林省教育厅科研项目(JJKH20230281KJ), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1205916736348549215, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, xref=null, ext=[AuthorCompanyExt(id=1205916736361132128, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, companyId=1205916736348549215, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Information and Control Engineering, Jilin Institute of Chemical Technology, Jilin 132022, China), AuthorCompanyExt(id=1205916736373715041, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, companyId=1205916736348549215, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=吉林化工学院信息与控制工程学院, 吉林 132022)])], figs=[ArticleFig(id=1205916738781245557, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, label=Fig.1, caption=The flow of the proposed model, figureFileSmall=4EwWc7XLAlhDAY4evzxV9Q==, figureFileBig=j4hgZo3Tgl1mp1i0Fl8PxA==, tableContent=null), ArticleFig(id=1205916738848354422, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, label=图1, caption=模型流程示意图, figureFileSmall=4EwWc7XLAlhDAY4evzxV9Q==, figureFileBig=j4hgZo3Tgl1mp1i0Fl8PxA==, tableContent=null), ArticleFig(id=1205916738961600631, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, label=Fig.2, caption=Accuracy comparison of different types of algorithms, figureFileSmall=JHxltEnl4+hlstpy09JwCA==, figureFileBig=dJ5qNAwE2dcsANf2hgMHDA==, tableContent=null), ArticleFig(id=1205916739024515192, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, label=图2, caption=不同类型算法准确率对比, figureFileSmall=JHxltEnl4+hlstpy09JwCA==, figureFileBig=dJ5qNAwE2dcsANf2hgMHDA==, tableContent=null), ArticleFig(id=1205916739091624057, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, label=Fig.3, caption=Key difference graph based on algorithm, figureFileSmall=CCspaosKVYP18tqA18B2xg==, figureFileBig=s1xEVG05Bj285WI5w+OJSA==, tableContent=null), ArticleFig(id=1205916739158732922, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, label=图3, caption=基于算法的关键差异图

水平连线表示同一分组内分类器之间没有显著差异

, figureFileSmall=CCspaosKVYP18tqA18B2xg==, figureFileBig=s1xEVG05Bj285WI5w+OJSA==, tableContent=null), ArticleFig(id=1205916739230036091, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, label=Fig.4, caption=Key difference graph based on data set, figureFileSmall=7juygmSfUstguv4ohH9HKw==, figureFileBig=402wadPvZ1xXU4zSbuacIg==, tableContent=null), ArticleFig(id=1205916739330699388, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, label=图4, caption=基于数据集的关键差异图

水平连线表示同一分组内分类器之间没有显著差异

, figureFileSmall=7juygmSfUstguv4ohH9HKw==, figureFileBig=402wadPvZ1xXU4zSbuacIg==, tableContent=null), ArticleFig(id=1205916739381031037, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, label=Table 1, caption=

UEA data set

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集 训练集 测试集 序列数 序列长度 类别数
AF 15 15 2 640 3
BM 40 40 6 100 4
Ep 137 138 3 206 4
EC 261 263 3 1 751 4
FM 316 100 28 50 2
HMD 160 74 10 400 4
Hb 204 205 61 405 2
RS 151 152 6 30 4
SRS1 268 293 6 896 2
SRS2 200 180 7 1 152 2
SWJ 12 15 4 2 500 3
), ArticleFig(id=1205916739456528510, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, label=表1, caption=

UEA数据集

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集 训练集 测试集 序列数 序列长度 类别数
AF 15 15 2 640 3
BM 40 40 6 100 4
Ep 137 138 3 206 4
EC 261 263 3 1 751 4
FM 316 100 28 50 2
HMD 160 74 10 400 4
Hb 204 205 61 405 2
RS 151 152 6 30 4
SRS1 268 293 6 896 2
SRS2 200 180 7 1 152 2
SWJ 12 15 4 2 500 3
), ArticleFig(id=1205916739527831679, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, label=Table 2, caption=

Comparison on accuracy of classification algorithms based on distance

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集 准确率/%
EDI DTWI DTWD SMTS
AF 0.167 0.267 0.200 1.000
BM 1.000 1.000 0.975 1.000
Ep 0.564 1.000 0.964 0.920
EC 0.361 0.323 0.844
FM 0.489 0.530 0.730
HMD 0.210 0.206 0.716
Hb 0.683 0.500 0.604 0.776
RS 0.869 0.891 0.818 0.921
SRS1 0.841 0.806 0.775 0.997
SRS2 0.447 0.489 0.539 1.000
SWJ 0.333 0.333 0.200 1.000
), ArticleFig(id=1205916739611717760, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, label=表2, caption=

基于距离的分类算法准确率对比

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集 准确率/%
EDI DTWI DTWD SMTS
AF 0.167 0.267 0.200 1.000
BM 1.000 1.000 0.975 1.000
Ep 0.564 1.000 0.964 0.920
EC 0.361 0.323 0.844
FM 0.489 0.530 0.730
HMD 0.210 0.206 0.716
Hb 0.683 0.500 0.604 0.776
RS 0.869 0.891 0.818 0.921
SRS1 0.841 0.806 0.775 0.997
SRS2 0.447 0.489 0.539 1.000
SWJ 0.333 0.333 0.200 1.000
), ArticleFig(id=1205916739683020929, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, label=Table 3, caption=

Comparison on accuracy of classification algorithms based on Shapelet

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集 准确率/%
ShapeNet gRFS Shapelet_D-S Multi-Shapelet SMTS
AF 0.167 0.267 0.550 0.500 1.000
BM 1.000 1.000 1.000 1.000 1.000
Ep 0.982 0.979 1.000 0.982 0.920
EC 0.346 0.725 0.844
FM 0.582 0.642 0.730
HMD 0.431 0.504 0.716
Hb 0.756 0.640 0.802 0.781 0.776
RS 0.875 0.891 0.935 0.918 0.921
SRS1 0.867 0.823 0.900 0.884 0.997
SRS2 0.789 0.517 0.643 0.868 1.000
SWJ 0.400 0.333 0.450 0.833 1.000
), ArticleFig(id=1205916739758518402, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, label=表3, caption=

基于Shapelet的分类算法准确率对比

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集 准确率/%
ShapeNet gRFS Shapelet_D-S Multi-Shapelet SMTS
AF 0.167 0.267 0.550 0.500 1.000
BM 1.000 1.000 1.000 1.000 1.000
Ep 0.982 0.979 1.000 0.982 0.920
EC 0.346 0.725 0.844
FM 0.582 0.642 0.730
HMD 0.431 0.504 0.716
Hb 0.756 0.640 0.802 0.781 0.776
RS 0.875 0.891 0.935 0.918 0.921
SRS1 0.867 0.823 0.900 0.884 0.997
SRS2 0.789 0.517 0.643 0.868 1.000
SWJ 0.400 0.333 0.450 0.833 1.000
), ArticleFig(id=1205916739829821571, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=EN, label=Table 4, caption=

Comparison of accuracy of classification algorithms based on network

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集 准确率/%
TapNet SMATE DA-Net MF-Net SMTS
AF 0.333 0.133 0.414 0.466 1.000
BM 1.000 1.000 0.925 0.950 1.000
HMD 0.365 0.527 0.347 0.445 0.716
Hb 0.727 0.727 0.626 0.692 0.776
SRS2 0.550 0.556 0.561 0.533 1.000
SWJ 0.400 0.200 0.400 0.400 1.000
), ArticleFig(id=1205916739892736132, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156908300735439743, language=CN, label=表4, caption=

基于网络的分类算法准确率对比

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集 准确率/%
TapNet SMATE DA-Net MF-Net SMTS
AF 0.333 0.133 0.414 0.466 1.000
BM 1.000 1.000 0.925 0.950 1.000
HMD 0.365 0.527 0.347 0.445 0.716
Hb 0.727 0.727 0.626 0.692 0.776
SRS2 0.550 0.556 0.561 0.533 1.000
SWJ 0.400 0.200 0.400 0.400 1.000
)], attaches=null, journal=Journal(id=1146119176004939786, delFlag=0, nameCn=科学技术与工程, nameEn=Science Technology and Engineering, nameHistory1=null, nameHistory2=null, issn=1671-1815, eissn=, cn=11-4688/T, coden=null, periodic=4, language=CN, oaType=是, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=UKU/O7GSka5polgCTkbIIw==, journalPrice=null, startedYear=null, abbrevIsoEn=Sci Technol Eng, journalRemark=null, publicationField=null, createdTime=null, updatedTime=1754445529766, createdBy=null, updatedBy=13701087609, firstLetterCn=S, firstLetterEn=S, subjectCode=Natural Sciences, subjectName=自然科学, subjectCodeEn=Natural Sciences, subjectNameEn=null, picCn=UKU/O7GSka5polgCTkbIIw==, picEn=5hwlULoNwcbj3xUmVi9MAQ==, jcr=null, cjcr=null, exts=[JournalExt(id=1159791870395564357, language=CN, name=科学技术与工程, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=null, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=http://www.stae.com.cn/jsygc/home, createdTime=1754445529793, updatedTime=1754445529793, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=http://www.stae.com.cn/jsygc/site/menus/20090429150146001, submissionAuthorUrl=http://www.stae.com.cn/jsygc/author/login, submissionEditorUrl=http://www.stae.com.cn/jsygc/editor/login, submissionReviewUrl=http://www.stae.com.cn/jsygc/reviewer/login, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1159791870441701702, language=EN, name=Science Technology and Engineering, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=null, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=http://www.stae.com.cn/jsygc/home, createdTime=1754445529804, updatedTime=1754445529804, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.stae.com.cn/jsygc/author/login, submissionEditorUrl=http://www.stae.com.cn/jsygc/editor/login, submissionReviewUrl=http://www.stae.com.cn/jsygc/reviewer/login, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1146123166801305609, websiteList=[Website(id=1148243202391400884, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1146123166801305609, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/kxjsygc/CN, language=CN, createTime=1751692112777, createBy=18614031015, updateTime=1753520965431, updateBy=18614031015, name=科学技术与工程-中文站点, tplId=1146099689490845704, title=科学技术与工程, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1148622798802673703, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=articleTextType, value=kx, createTime=1751782615614, updateTime=1751782615614, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798781702180, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=banner, value=null, createTime=1751782615609, updateTime=1751782615609, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798769119267, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=logo, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic?fileId=j86gbwi+p0Idkyl5SzIlmQ==, createTime=1751782615606, updateTime=1751782615606, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798794285094, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic, createTime=1751782615612, updateTime=1751782615612, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798790090789, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1751782615611, updateTime=1751782615611, creator=18614031015, updator=18614031015)]), Website(id=1155914124811976731, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1146123166801305609, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/kxjsygc/EN, language=EN, createTime=1753521003206, createBy=18614031015, updateTime=1753521003206, updateBy=18614031015, name=科学技术与工程-英文站点, tplId=1146101810881728533, title=Science Technology and Engineering, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1155914371227308235, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=articleTextType, value=kx, createTime=1753521061952, updateTime=1753521061952, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371210531016, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=banner, value=null, createTime=1753521061947, updateTime=1753521061947, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371202142407, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=logo, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic?fileId=j86gbwi+p0Idkyl5SzIlmQ==, createTime=1753521061945, updateTime=1753521061945, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371223113930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic, createTime=1753521061950, updateTime=1753521061950, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371218919625, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1753521061949, updateTime=1753521061949, creator=18614031015, updator=18614031015)])], journalTitle=科学技术与工程, weixinUrl=null, journalUrl=null, iacademicId=null, status=0, seqNo=null, journalTitleEn=Science Technology and Engineering, journalPhotoCn=UKU/O7GSka5polgCTkbIIw==, journalPhotoEn=5hwlULoNwcbj3xUmVi9MAQ==, journalFirstLetter=S, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=null, provinceCode=null, provinceName=null, collectFlag=false), detailUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2308252, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2308252, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2308252, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2308252, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于Shapelets的多元时间序列分类方法
收藏切换
PDF下载
王威娜 , 李明莉
科学技术与工程 | 论文·自动化技术、计算机技术 2025,25(1): 252-261
收起
收藏切换
科学技术与工程 | 论文·自动化技术、计算机技术 2025, 25(1): 252-261
基于Shapelets的多元时间序列分类方法
全屏
王威娜 , 李明莉
作者信息
  • 吉林化工学院信息与控制工程学院, 吉林 132022
  • 王威娜(1981—),女,汉族,吉林吉林人,博士,教授。研究方向:时间序列分析、数据挖掘。E-mail:

Multivariate Time Series Classification Method Based on Shapelets
Wei-na WANG , Ming-li LI
Affiliations
  • School of Information and Control Engineering, Jilin Institute of Chemical Technology, Jilin 132022, China
出版时间: 2025-01-08 doi: 10.12404/j.issn.1671-1815.2308252
文章导航
收藏切换

多元时间序列分类是众多领域的关键问题,但是当前多元时序分类研究面临着原始数据高维、精度不足、可解释性缺乏等问题,这使得模型性能提升受限,准确率难以满足实际需求。针对上述问题,提出基于Shapelets的多元时间序列分类方法。首先,利用自适应邻居的无监督Shapelet学习将Shapelet变换与自适应权重结合,用于自动学习显著多元Shapelets;然后,将该方法与Shapelet相似性和类标约束项结合,增强模型可解释性和分类准确性;最后,提出模型的优化策略,用以获取最优的Shapelets,进一步提高模型的分类精度。与3种不同类型11个算法在11个公开数据集上进行比较,实验结果表明提出算法具有较高的分类精度。

多元时间序列  /  多元时间序列分类  /  Shapelets学习  /  优化策略

Multivariate time series classification is a key problem in many fields, but the current research on multivariate time series classification is faced with some problems, such as high dimensionality of original data, low accuracy, and lack of interpretability, which limits the performance improvement of models and makes it difficult to meet the actual requirements. Aiming at above problem, a multivariate time series classification method based on Shapelets was proposed. Firstly, unsupervised Shapelet learning of adaptive neighbors was used to automatically learn significant multivariate Shapelets by combining Shapelets transform and adaptive weights. Then, the method was combined with Shapelet similarity and class label constraint to enhance the interpretability and classification accuracy of the model. Finally, the optimization strategy of the model was proposed to obtain the best Shapelets to further improve the classification accuracy of the model. Three different types of 11 algorithms were compared on 11 public data sets, and the experimental results show that the proposed algorithm has high classification accuracy.

multivariate time series  /  multivariate time series classification  /  Shapelets learning  /  optimization strategy
王威娜, 李明莉. 基于Shapelets的多元时间序列分类方法. 科学技术与工程, 2025 , 25 (1) : 252 -261 . DOI: 10.12404/j.issn.1671-1815.2308252
Wei-na WANG, Ming-li LI. Multivariate Time Series Classification Method Based on Shapelets[J]. Science Technology and Engineering, 2025 , 25 (1) : 252 -261 . DOI: 10.12404/j.issn.1671-1815.2308252
随着人工智能的发展,时间序列数据的数量呈直线上升趋势,对其分析与研究已引起学术界和工业界的广泛关注[1]。早期研究者主要关注一元时间序列分类问题,但由于描述对象单一,导致其在众多领域的应用存在局限性。因此,越来越多的科研工作者开始深入研究多元时间序列[2]。多元时间序列具有广泛的实际应用,例如生物医学诊断[3]、股票金融市场[4]、交通流量[5]、工业制造[6]、人类活动识别[7]等。多元时间序列可以被看作是多个一元时间序列的集合,在整个学习过程中不仅要对一元时间序列进行分析,还需要充分考虑各变量之间的相关性。以自动驾驶为例,为做出更精准的决策,需分析多种传感器信息,包括摄像头和激光雷达等,并将其综合起来进行全面决策[8]。多元时间序列分类的研究主要包括基于距离的方法[9]、基于特征的方法[10]、基于深度学习的方法[11]和基于模型的方法[12]。基于距离的方法主要通过计算不同时间序列之间的距离进行分类,同类间距离相近而不同类间距离较远,常用方法包括欧几里得距离、基于概率的距离、动态时间规整(dynamic time warping, DTW)距离[13]。动态时间规整距离已在多元时间序列分类领域展现出卓越性能,但是该方法的计算过程十分烦琐。基于特征的方法通过从原始时间序列数据中提取相关特征,并将这些特征作为输入进行分类,主流方法包括二维奇异值分解、时间序列符号表示[14]、Shapelets方法[15-17]。在基于深度学习的方法中,循环神经网络[18]、卷积神经网络[19]等方法已成为多元时序分类领域备受瞩目的研究对象,这些方法通过构建模型学习时间序列数据的内部关系和结构,提取特征实现分类,在分类准确度方面具有明显优势,但是可解释性较弱[20]。基于模型的方法将原始多元时间序列实例转换为模型参数,建立数学模型获取时间序列数据特征,并利用该模型进行分类。与其他方法相比,基于Shapelets的分类方法越来越受到科研学者的重视[21],这是因为此类方法可以高效地获取信息,并具有较高的可解释性和准确度。
Shapelets是指时间序列中最具辨识性的子序列,这一概念由Keogh等[22]于2009年首次提出。近年来,Shapelets已成为备受关注的研究热点。Shapelets具有很好的可解释性,能够有效地区分时间序列数据中的类间差异,另外Shapelets是从时间序列中选取的最具有代表性的特征,同时也达到了对时序数据降维的效果,但Shapelets的选取过程具有较高的时间复杂度,限制了模型的发展与应用[23]。大量快速发现Shapelets的算法被提出,此类方法通过信息增益度量的熵修剪方法替换传统欧几里得距离的计算,后续也有研究者提出搜索空间的修剪方法,用以提升Shapelets的发现速度[24]。Lines等[25]利用最优的Shapelets与时间序列之间的距离作为时间序列的新特征,并结合其他分类器实施最终分类,为分类算法的发展提供了新思路。Grabocka等[26]提出学习时间序列Shapelets算法(learning time-series Shapelets, LTS),该算法基于特定目标函数进行优化Shapelets,结果表明与其他Shapelets的分类器相比,LTS算法的准确率得到显著提高。最近,Guillaume等[27]引入扩张Shapelet变换(dilated Shapelet transform, DST),获得包含膨胀概念的Shapelet,并利用Shapelet特征增强分类能力, 有效扩展了Shapelets的应用领域。Chen等[28]提出可解释时序分类的小波选择方法,通过定义位置度量和距离度量评价每个候选Shapelet的判别能力,从而在分类结果中提供更具可解释性的洞察力,但在筛选Shapelet的过程中,此方法的计算复杂度较高。Liu等[29]提出基于具有规范时序特征Shapelet的时序分类算法,利用提取的Shapelet建立新的特征表示,构建随机森林分类器,保证算法的泛化能力。虽然将Shapelet应用于时序分类的研究已取得长足发展,但是仍面临着Shapelet选取过程烦琐和显著性欠缺的挑战。王威娜等[30]提出基于优化Shapelet的时序分类方法,虽然选取Shapelet过程简便且具有可解释性,但是只考虑了单变量时间序列,对多变量问题未深入研究。李晨等[31]提出基于优化和两阶段筛选的Shapelets提取方法,虽然能够获取优异的Shapelets集合,但是后续筛选过程的复杂度较高。杨骏等[32]提出面向时序有序分类的Shapelet抽取算法,结合SAX与Shapelet,并通过布隆过滤器对候选Shapelet进行剪枝,但该过程也相对烦琐。
针对上述问题,现提出基于Shapelets的多元时间序列分类方法(classification method based on Shapelets on multivariate time series, SMTS)。首先,利用自适应邻居的无监督Shapelet学习方法,将Shapelet变换与自适应权重结合用于自动学习显著多元Shapelets,并根据自适应权重的计算,获取Shapelets的显著性。然后,将Shapelet学习方法与Shapelets相似性和类标约束项结合,计算Shapelets之间的相似程度,此过程与类标约束项结合,能够更好地减小类标之间的差异,缩小模型误差。最后,提出优化策略获取最优且不等长的多元Shapelets,进而以此为基础实现多元时间序列分类。该算法通过自适应权重的无监督学习方法突出Shapelets的显著性,并融合Shapelets的多样性、相似性和类标约束,优化Shapelets的选取过程,减少分类复杂度的同时提高分类的准确性。
列出的定义分别为时间序列、Shapelet、时间序列和Shapelet之间的距离度量以及Shapelet变换技术。
定义1 (时间序列) 时间序列是一组序列数据,它由n个数值型数据组成,一条完整的时间序列可表示为t={x1,x2,…,xn},其中x1,x2,…,xn按照时间顺序排列。多条时间序列可以组成一个时间序列集,即T={t1,t2,…,tm}。多元时间序列数据中包含l个时间序列,即MTS={MTS1, MTS2,…,MTSl},每个时间序列为MTSi R P × Q,P为每个样本包含的一元时间序列的个数,Q为一维样本的长度。
定义2 (Shapelet) Shapelet是时间序列中特征最显著的一段子序列。在一元时间序列分类中,选出K个最优的Shapelet,并记为S R K × M,M为时间序列的长度,即S={S1,S2,…,SK}。
定义3 (时间序列和Shapelet之间的距离) 一条时间序列t与第k个Shapelet Sk之间的距离如式 (1)所示。它通过滑动窗口,依次计算Sk与时间序列所有子序列之间的欧式距离,并选取最短距离作为最终结果。
$\operatorname{dist}=\min _{q=1,2, \cdots, \bar{q}}\left[\frac{1}{l_{k}} \sum_{k=1}^{l_{k}}\left(t_{q+k-1}-s_{k}\right)^{2}\right]$
定义4 (Shapelet变换技术) 利用定义3可将时间序列和Shapelet之间的距离作为新的特征,把时间序列映射到新特征空间中的过程称为Shapelet变换技术[29],经Shapelet变换之后数据集转换为D,记为D={D1,D2,…,Dn},其中 D i ={Di,1,Di,2,…,Di,K},由于K<Q,所以在保留时间序列关键信息的同时实现了数据的降维处理。
提出基于Shapelets的多元时间序列分类模型,该模型由自适应邻居的无监督Shapelet学习及Shapelet最小相似性和类标约束项共同构建。首先,自适应邻居无监督Shapelet学习方法可自动学习时序数据显著信息,进而获取多元Shapelets;然后,结合Shapelet最小相似性和类标约束项构建候选Shapelets模型;最后,优化更新Shapelets、分类器和稀疏邻居矩阵直到其收敛,从而获得最优的Shapelets,并以此为基础实施多元时序数据分类。模型的具体流程如图1所示。
Shapelet学习是解决时间序列分类问题的关键技术之一[33]。自适应邻居的无监督Shapelet学习方法可以有效解决不同序列的重要度计算和多元Shapelets的更新策略[34]中存在的问题,该方法可以自动学习多元Shapelets信息,当给定候选多元Shapelets时,可以自动确定每个信息变量的显著性。
给定候选的多元Shapelets,可将多元时间序列转换为基于多元Shapelets的矩阵X=[Xij],表达式为
$\begin{aligned} X_{i j} & =\operatorname{dist}\left[t^{(i)}, s^{(j)}\right] \\ & =\min _{q=1,2, \cdots, \bar{q}} \sum_{f=1}^{F} \beta_{k}^{f}\left\{\frac{1}{l_{k}} \sum_{h=1}^{l_{k}}\left[t_{q+k-1}^{f(i)}-s_{k}^{f(j)}\right]^{2}\right\} \\ & =\min _{q=1,2, \cdots, \bar{q}} \sum_{f=1}^{F} \beta_{k}^{f} d_{i j q}^{f} \end{aligned}$
式(2)中:Xij为第i个多元时间序列与第j个多元Shapelets之间的距离; q -=qi-lj+1为时间序列ti与Shapelet Sj之间的总段数;qi为多元时间序列的长度;lj为多元Shapelets的长度;F为多元时间序列中变量的个数;β为不同变量的重要性权重; t q + k - 1 f ( i )为在第f个一元时间序列中长度为(q+k-1)的第i条时间序列; s k f ( j )为在第f个Shapelet中长度为k的第j行Shapelet。利用最小函数近似表示X=[Xij]为
$\begin{aligned} X_{i j} & =\min _{q=1,2, \cdots, \bar{q}} \sum_{f=1}^{F} \beta_{k}^{f} d_{i j q}^{f} \\ & \approx \frac{\sum_{q=1}^{\bar{q}}\left[\sum_{f} \beta_{k}^{f} d_{i j q}^{f} \exp \left(\alpha \sum_{f} \beta_{k}^{f} d_{i j q}^{f}\right)\right]}{\sum_{q=1}^{\bar{q}} \exp \left(\alpha \sum_{f} \beta_{k}^{f} d_{i j q}^{f}\right)} \\ & =\frac{\sum_{f=1}^{F} \beta_{k}^{f}\left[\sum_{q=1}^{\bar{q}} d_{i j q}^{f} \exp \left(\alpha \sum_{f} \beta_{k}^{f} d_{i j q}^{f}\right)\right]}{\sum_{q=1}^{\bar{q}} \exp \left(\alpha \sum_{f} \beta_{k}^{f} d_{i j q}^{f}\right)} \\ & =\sum_{f=1}^{F} \beta_{k}^{f} F X_{i j}^{f} \end{aligned}$
式(3)中:
$F X_{i j}^{f}=\frac{\sum_{q=1}^{\bar{q}} d_{i j q}^{f} \exp \left(\alpha \sum_{f} \beta_{k}^{f} d_{i j q}^{f}\right)}{\sum_{q=1}^{\bar{q}} \exp \left(\alpha \sum_{f} \beta_{k}^{f} d_{i j q}^{f}\right)}$
式(4)中:α用以控制函数的精度,根据前人的工作经验[35],将α设置为-100;令XkXij构成的矩阵,F X k f= { F X i j f } j = 1 J表示第f个多元时间序列与第i个多元Shapelets之间的距离。
针对不同变量重要性的差异,学习方法可自适应确定不同变量的重要性权重。权重βk包括测量 β k LFXk之间一致性的可学习权重 β k L和多个变量之间相等的固定权重 β k F,其中βk的学习过程可以看作是一种平行注意机制。另外,βk的计算可以表示为
βk=(1-μ) β k L+μ β k F=(1-μ) e x p ( b k ) f e x p ( b k f )+μ 1 F
bk=bk+ X T kFXk
X k= X k 1 + X k 2Xk
式中:μ为权重参数,取值为μ∈[0,1];bk通过度量XkFXk的一致性来迭代细化。
拉普拉斯映射在无监督学习中被广泛采用[30],其作用是使得相似时间序列尽可能具有相同的伪类标签。G R n × n表示经过Shapelet变换的时间序列相似矩阵,即
$\boldsymbol{G}_{(i j)}=\exp \left[-\frac{\left\|\boldsymbol{X}_{(:, i)}-\boldsymbol{X}_{(: j)}\right\|^{2}}{\sigma^{2}}\right]$
式(8)中:σ为RBF内核参数;X(:,i)X(:,j)分别为X矩阵中第i列的子矩阵和第j列的子矩阵。根据获得的G,相应的拉普拉斯正则化项可表示为
1 2 i = 1 n j = 1 n G(ij) = Y ( : , i ) - Y ( : , j ) = 2 2= 1 2 k = 1 c i = 1 n j = 1 n G(ij) [ Y ( k i ) - Y ( k j ) ] 2= k = 1 c Y(k,:)(DG-G) Y T ( k , : )=tr(YLGYT)
式(9)中:DG(i,i)= j = 1 n G(ij)为对角矩阵;LG=DG-G
为保障Shapelets的多样性,尽量剔除相似特性的Shapelets,从而增加Shapelets相似性模块,使得输出的Shapelets具有鲜明特色。假设Shapelets相似矩阵为H R k × k,Shapelets SmSn之间的相似性矩阵元素 H ( m n )
H(mn)= e - d m n 2 σ 2
式(10)中:dmn为Shapelet的SmSn之间的距离,计算方法与式(2)相似。
模型中设立两项类标约束项,其目的是减少生成类标和实例类标之间的差异性,增强类标分配的准确性。
第一项是通过F范数实现计算实例类标和生成类标之间误差的平方和,获得规范化模型参数的同时也能使误差最小化,利用式(11)所示的函数实现误差最小化,表达式为
m i n W = W T X - Y = F 2
第二项是计算生成类标与变换矩阵转置的乘积以及实际类标和变换矩阵转置的乘积,将得到的前后两个结果相减,通过F范数获得误差最小值,即
m i n Y = Y D T - ( Y - I ) D T = F 2= m i n Y = ( 2 Y - I ) D T = F 2
式(12)中:D R k × n为Shapelet变换矩阵;Y∈Rc×n为类标约束矩阵;I为单位矩阵;W∈Rk×c为分类器矩阵。
当给定候选的多元Shapelets S后,SMTS可以获得基于多元Shapelets的表示X(S)。为了学习信息丰富的多元Shapelets,SMTS同时学习邻居分配矩阵A、分类器W和候选多元Shapelets S。为保证获取的多元Shapelets能够代表多元时间序列的不同特征,构造目标函数包含6个主要成分。
第一部分$\left[\sum_{i, j=1}^{n}\left\|\boldsymbol{X}(\boldsymbol{S})_{. i}-\boldsymbol{X}(\boldsymbol{S})_{. j}\right\|_{F}^{2} \boldsymbol{A}_{i j}\right]$为邻居一致性项,它能更准确地表示时间序列之间的相似性,并且采用的最优邻居可提高算法的效率。
第二部分(‖A2)为邻居正则化项,优化模型的损失和复杂度,减少过拟合,提高模型的泛化能力。
第三部分{tr[YL(A)YT]}为拉普拉斯正则化项,确保多元时间序列数据之间的局部相似性,并提高模型的可解释性。
第四部分[‖H(S)‖2]为多元Shapelet相似性项,用以保证Shapelet的多样化,同时也会使得分类器能够更好地区分不同类别的时间序列,提高模型的分类精度。
第五部分(‖W F 2)为最优分类器,因为它可以更精确地匹配不同类别的时间序列样本,用以提高分类模型的准确性;
第六部分[‖WTD(S)-Y F 2和‖(2Y-ID(S)T F 2]为两项类标约束项,对分类器进行正则化和约束,使其更加符合数据的分布,这样有助于减小过拟合的风险,使得数据之间的差异最小化,确保分类任务的性能。
因此,SMTS的目标函数可以表示为
$\begin{array}{l} \min _{A, Y, S} \frac{1}{2} \sum_{i=1}^{n} \sum_{j=1}^{n}\left\|\boldsymbol{X}(\boldsymbol{S})_{. i}-\boldsymbol{X}(\boldsymbol{S})_{. j}\right\|_{\mathrm{F}}^{2} \boldsymbol{A}_{i j}+ \\ \frac{\gamma}{2}\|\boldsymbol{A}\|^{2}+\lambda_{0} \operatorname{tr}\left[\boldsymbol{Y} \boldsymbol{L}(\boldsymbol{A}) \boldsymbol{Y}^{\mathrm{T}}\right]+\frac{\lambda_{1}}{2}\|\boldsymbol{H}(\boldsymbol{S})\|_{F}^{2}+ \\ \frac{\lambda_{2}}{2}\|\boldsymbol{W}\|_{F}^{2}+\frac{\lambda_{3}}{2}\left\|\boldsymbol{W}^{\mathrm{T}} \boldsymbol{D}(\boldsymbol{S})-\boldsymbol{Y}\right\|_{F}^{2}+ \\ \quad \frac{\lambda_{4}}{2}\left\|(2 \boldsymbol{Y}-\boldsymbol{I}) \boldsymbol{D}(\boldsymbol{S})^{\mathrm{T}}\right\|_{F}^{2} \\ \text { s.t. } \rightarrow \boldsymbol{A}_{i}^{\mathrm{T}} 1=1,0 \leqslant \boldsymbol{A}_{i j} \leqslant 1, \boldsymbol{Y} \boldsymbol{Y}^{\mathrm{T}}=\boldsymbol{I} \end{array}$
采用坐标下降法求解SMTS目标函数,具体方法如下。
当固定WS时,SMTS的目标函数退化为
m i n AF(A)= 1 2 i , j = 1 N = X ( S ) . i - X S . j = F 2Aij+ γ 2 = A = 2+λ0tr[YL(A)YT]s.t.→ A T i1=1,0≤Aij≤1,YYT=I
利用谱分析中的基本方程得
tr(YLAYT)= 1 2 i = 1 n j = 1 n = Y . i - Y . j = F 2Aij
则上述函数F(A)可表示为
m i n A T i I = 1,0 A i j 1 j = 1 n = X ( S ) . i - X ( S ) . j = F 2Aij+γ = A = 2+λ0 j = 1 n = Y . i - Y . j = F 2Aij m i n A T i I = 1,0 A i j 1 j = 1 n = 1 2 γ d i A + A i j = 2 m i n A i , η , β iF(Ai,ξ,βi)= 1 2 = A i + d i 2 γ = F 2-ξ( A T iI-1)- β T iAi
式(16)中:η>0、βi>0为拉格朗日乘数;I为全为1的列向量。由于A是稀疏的,因此只有与X(S ) . i最近的k个邻居才可能连接到X(S ) . i。然后,确定F(Ai, ξ, βi)的最优解为
$\boldsymbol{A}_{i}=\max \left(\xi-\frac{d_{i}}{2 \gamma}, 0\right)$
Ai k ¯个非零元素,其中
ξ= 1 k ¯+ 1 2 k ¯ γ j = 1 k ¯ dij
γ= 1 N i = 1 N k ¯ 2 d i , k ¯ + 1 - 1 2 j = 1 k ¯ d i j
当固定AW时,SMTS的目标函数退化为
m i n SF(S)= i , j = 1 n = X ( S ) . i - X S . j = F 2Aij+ λ 1 2 = H ( S ) = F 2+ λ 3 2 = W T D ( S ) - Y = F 2+ λ 4 2 = ( 2 Y - I ) D ( S ) T = F 2
采用迭代算法更新S,即$\boldsymbol{S}_{i+1}=\boldsymbol{S}_{i}-\eta \nabla \boldsymbol{S}_{i}$,其中η为学习率。F(S)对S的导数为
$\begin{aligned} \frac{\partial F(\boldsymbol{S})}{\partial \boldsymbol{s}_{k p}^{f}}= & \lambda_{1} \boldsymbol{H}(\boldsymbol{S}) \frac{\partial \boldsymbol{H}(\boldsymbol{S})}{\partial \boldsymbol{s}_{k p}^{f}}+ \\ & \sum_{i, j}\left\{\boldsymbol { A } _ { i j } [ \boldsymbol { X } ( \boldsymbol { S } ) _ { k i } - \boldsymbol { X } ( \boldsymbol { S } ) _ { k j } ] \left[\frac{\partial \boldsymbol{X}(\boldsymbol{S})_{k i}}{\partial \boldsymbol{s}_{k p}^{f}}-\right.\right. \\ & \left.\left.\frac{\partial \boldsymbol{X}(\boldsymbol{S})_{k j}}{\partial \boldsymbol{s}_{k p}^{f}}\right]\right\}+\lambda_{3}\left(\boldsymbol{W}^{\mathrm{T}} \boldsymbol{D}-\boldsymbol{Y}\right) \frac{\partial \boldsymbol{D}(\boldsymbol{S})}{\partial \boldsymbol{s}_{k p}^{f}}+ \\ & \lambda_{4}(2 \boldsymbol{Y}-\boldsymbol{I}) \frac{\partial \boldsymbol{D}(\boldsymbol{S})}{\partial \boldsymbol{s}_{k p}^{f}} \end{aligned}$
此外,式(21)第一项中的 H ( S ) i j S k p f变为
H ( S ) i j s k p f=- 2 σ 2H(S)ij d i j S d i j S s k p f
式(22)中: d i j SSiSj之间的距离; d i j S d i j S S k p f的计算方法与X(S)ij X ( S ) i j S k p f相同。式(21)的第二项 X ( S ) i j S k p f变为
$\begin{aligned} \frac{\partial \boldsymbol{X}(\boldsymbol{S})_{i j}}{\partial \boldsymbol{s}_{k p}^{m}}= & \frac{1}{E_{2}^{2}} \sum_{q=1}^{\bar{q}} \exp \left(\alpha \sum_{f^{\prime}} \beta_{k}^{f^{\prime}} d_{i j q}^{f^{\prime}}\right) \times \\ & {\left.\left[1+\alpha \beta_{k}^{f} d_{k n q} f_{k}^{\prime}\right) E_{1}-\alpha E_{2}\right] \beta_{k}^{f} \frac{\partial d_{i j q}^{f}}{\partial s_{k p}^{f}} } \end{aligned}$
$E_{1}=\sum_{q=1}^{\bar{q}} \beta_{k}^{f} d_{i j q}^{f} \exp \left(\alpha \sum_{f^{\prime}} \beta_{k}^{f^{\prime}} d_{i j q}^{f^{\prime}}\right)$
$E_{2}=\sum_{q=1}^{\bar{q}} \exp \left(\alpha \sum_{v} \beta_{k}^{f^{\prime}} d_{i j q}^{\prime}\right)$
式中:
d i j q f S k p f= 2 l k[ S p f ( i )-m t q + p + 1 f ( j )], q -=qi-lj+1。
综上,$\nabla \boldsymbol{S}=\frac{\partial F(\boldsymbol{S})}{\partial \boldsymbol{S}}$可由式(21)~式(25)共同计算得出。
当固定AS时,SMTA的目标函数退化为
$\min _{W} F(\boldsymbol{W})=\min _{W}\left[\frac{\lambda_{2}}{2}\|\boldsymbol{W}\|_{F}^{2}+\frac{\lambda_{3}}{2}\left\|\boldsymbol{W}^{\mathrm{T}} \boldsymbol{D}(\boldsymbol{S})-\boldsymbol{Y}\right\|_{F}^{2}\right]$
然后求解W的最小值,对F(W)关于W进行求导计算,公式为
F ( W ) W=λ2W+λ3D(WTD-Y)T=(λ2I+λ3DDT)W-λ3DYT
令其等于0,得到W的最终结果为
Wt+1= ( λ 2 I + λ 3 D t D T t ) - 1(λ3Dt Y T t)
提出的多元时序分类算法首先给出自适应邻居的Shapelet学习方法,计算时间序列和Shapelet之间的距离,并将其映射到Shapelet空间中,使得获取的Shapelet更精确,保证了可解释性的同时也提高了算法的效率;然后,结合Shapelet相似性和类标约束项,从而获得信息更丰富的Shapelets,同时减少了数据之间的差异,使其更加符合数据的分布,这样有助于减小过拟合的风险,提高模型的稳定性和泛化能力;最后,利用坐标下降法进行优化迭代更新,从而得到最优的Shapelets,并以此为基础实施分类。提出的分类算法的伪代码如算法1所示。
算法1 基于Shapelet的多元时序数据分类算法
输入:多元时间序列T,不同长度的多元Shapelets,迭代次数imax,学习比率η,参数λ0,λ1,λ2,λ3,λ4,α,μ
输出:多元时间序列的分类结果Tl
1. 初始化S,Y,W;
2. 根据式(3)~式(7),计算TS之间的距离,并映射到Shapelet空间中得到X(S);
3. 根据S和式(10)计算H(S) ;
4. 通过式(16)~式(17)更新A;
5. 通过式(28)更新W;
6. For i=1,2,…, i m a x  do
7. 通过式(21)计算$\nabla S_{i}$;
8. 通过公式$\boldsymbol{S}_{i+1}=\boldsymbol{S}_{i}-\eta \nabla \boldsymbol{S}_{i}$来更新Si;
9. End for
10. Sbest= S i m a x;(Sbest为获得的最佳Shapelets)
11. 计算Sbest与时间序列的距离并进行排序,确定Sbest的类标,获得带类标的Shapelets Sl;
12. 将得到的Sbest与时间序列的距离按照类别数进行排序,并根据Sl确定最终时间序列的类标,获得带有分类标签的时间序列Tl;
Return Tl
如算法1所述,每次迭代计算矩阵X(S)和H(S)的计算复杂度分别为O(MQNKL)和O(M2L2K2),其中M为多元时间序列中的变量数,Q为每个时间序列的最大长度,L为每个Shapelet的最大长度,N为指多元时间序列的大小,KS的最大值。针对SMTS的学习过程中后续AWS的依次迭代更新,通过固定S迭代更新A,W直到其收敛,更新AW的计算复杂性分别是O(N2K+N2C)和O(N2K+K3+NKC+K2C),其中C为类数。当更新AW直到收敛时,复杂度为O{J[N(NC+K2+NK+KC)+K2(K+C)]},J为迭代的最大次数。当更新S直到收敛时,复杂度为O[ i m a x  (M2Q2N2+M2L2K2+N2K)],imax为最大的迭代次数。综上所述,总复杂度为O(G{MQNKL+M2L2K2+ J[N(NC+K2+NK+KC)+K2(K+C)]+imax(M2Q2N+ M2L2K2+N2K)}),其中G是SMTS的最大迭代次数。因为K,CN,所以最终SMTS的时间复杂度可简化成O[G(M2Q2N2imax)]。
实验采用3 GHz和16 GB内存的CPU,Window 10操作系统,编程环境为MATLAB2021。提出算法在11个数据集上分别与3种不同类型(11个)分类算法进行对比,即基于距离的分类算法[32]、基于Shapelet的分类算法[33-36]、基于网络的分类算法[37-40],以此验证本文算法的有效性。
利用11个真实的多元时间序列数据集[41]评估所提出的模型,数据集分别为Atrial Fibrillation(AF)、Basic Motions(BM)、Epilepsy(Ep)、Ethanol Concentration(EC)、Finger Movements(FM)、Hand Movement Direction(HMD)、Racket Sports(RS)、Heartbeat(Hb)、Self Regulation SCP1(SRS1)、Stand Walk Jump(SWJ)、Self Regulation SCP2(SRS2),具体信息如表1所示。数据的维度为2~61,长度为30~2 500。对部分数据集进行了预处理,并利用训练集调整参数,利用测试集进行评估。
有两种不同类型的多元Shapelets:相同长度的多元Shapelets和不同长度的多元Shapelets。在实验中,正则化参数为λ0λ1λ2λ3λ4,从{10-4, 10-2, 100, 102, 104}中确定,参数μ从{0, 0.5, 1}中选取,内部迭代次数imax设为50,学习率设为0.01。
提出算法与3种基于距离的算法EDI、DTWI DTW D 36,在UEA的11个多元时间序列数据集上进行比较,实验对比结果如表2所示。表2表明提出算法在除Ep以外的10个数据集上得到最优的分类结果,且在AF、BM等数据集上均得到100%的准确率。
提出算法仅在Ep上稍有欠缺,因为该数据集的规律性不强,每个类别的个数不一致。但是整体的实验结果表明,在基于距离的分类算法中提出算法在准确度方面具有明显优势。
提出算法与基于Shapelet的算法ShapeNet[37],gRFS[38],Shapelet_D-S[39],Multi-shapelet[40]进行比较的结果如表3所示。可以看出,提出算法在大部分数据集均取得最好的结果。在Hb和RS数据集上,Shapelet_D-S算法仅以微小的优势胜过本文算法,但是从整体上看,利用本文方法获得的Shapelet集合具有显著性,使得时间序列基于Shapelet分类算法的结果在这11个数据集上具有明显优势。
提出算法与基于网络的分类算法TapNet[42],SMATE[43],DA-Net[44],MF-Net[45]准确率对比的结果如表4所示,可以看出,本文算法的准确率在6个数据集上均为最高,在4个数据集上达到100%的准确率,在与最先进的MF-Net算法进行对比时,提出算法也以明显优势获胜,这不仅表明提出算法的优越性,同时进一步显示出本文算法有效地提高了时间序列分类精度。
根据不同类型算法准确率列出统计图如图2所示。条形柱代表在同种类型中不同算法在11个数据集上胜出的个数。可清楚看出,本文算法在准确率上具有优势,在3种类型中均取得了突出成果。
进一步,对算法进行了Nemenyi非参数统计检验,并绘制了关键差异图,结果如图3图4所示。首先,将所有算法在6个公共数据集上进行Nemenyi非参数统计检验并绘制关键差异图,结果如图3所示。可以看出,本文算法的性能与其他分类器相比达到了最高的准确率。然后,将基于距离的分类算法和基于Shapelet的分类算法在11个数据集上进行统计检验,得到的关键差异图如图4所示。图4表明提出算法的分类结果最优,体现了提出算法的先进性和优越性,在与Shapelet-D_S的对比中也以0.1的优势获胜。
提出基于Shapelets的多元时间序列分类方法。首先,利用自适应邻居的无监督Shapelet学习方法将Shapelet变换和自适应权重结合,并用于自动学习显著多元Shapelets。通过自适应权重的计算,得到不同Shapelet的重要性,进而提高模型的分类精度。然后,该方法与Shapelet相似性和类标约束项相结合,通过Shapelet相似性可以增强模型可解释性,类标约束项使得实际类标和生成类标的差异最小化,提高分类的准确性。最后,提出模型的优化策略来获取最优的且不同长度的多元Shapelets。
将提出算法在UEA的11个数据集上进行验证,并且分别针对3种不同类型11个算法进行比对,得到以下结论。
(1) 在与基于距离的3个算法进行比较时,提出算法的优势十分明显。基于Shapelet分类也会对数据进行降维,减少了分类的复杂度,同时也提高了分类的精度。
(2) 在与基于Shapelet的4个算法进行对比时,提出算法在大部分数据集上的结果均为最佳,整体上优于其他先进算法的分类效果,这体现了本文算法的先进性和有效性。
(3) 在与基于网络的4个算法进行对比时,提出算法在6个数据集上均为最佳结果,甚至在4个数据集上均取得100%的准确率,展现了本文算法的显著优势。
准确率对比图和关键差异图进一步表明本文算法的优越性。本文算法相比于其他3种类型的算法,在分类精度上取得了明显的优势,同时也表明在获取Shapelet集合的可解释性和分类准确性的效果十分显著。未来将探索更高层次的Shapelet集合,使得模型获得更好的效率和分类精度。
  • 国家自然科学基金(62266046)
  • 吉林省自然科学基金(YDZJ202201ZYTS603)
  • 吉林省教育厅科研项目(JJKH20230281KJ)
参考文献 引证文献
排序方式:
[1]
Feremans L, Cule B, Goethals B. PETSC: pattern-based embedding for time series classification[J]. Data Mining and Knowledge Discovery, 2022, 36(3): 1015-1061.
[2]
张旭, 张亮, 金博, 等. 基于不确定性的多元时间序列分类算法研究[J]. 自动化学报, 2023, 49(4): 790-804.
Zhang Xu, Zhang Liang, Jin Bo, et al. Research on classification algorithm of multivariate time series based on uncertainty[J]. Acta Automatica Sinica, 2023, 49(4): 790-804.
[3]
Velasco I, Sipols A, De Blas C S, et al. Motor imagery EEG signal classification with a multivariate time series approach[J]. BioMedical Engineering OnLine, 2023, 22(1): 1-24.
[4]
Lee J, Youn H L, Poon J, et al. StockEmotions: discover investor emotions for financial sentiment analysis and multivariate time series[J]. ArXiv Preprint ArXiv, 2023: 2301.09279.
[5]
Yan Y, Zhang S, Tang J, et al. Understanding characteristics in multivariate traffic flow time series from complex network structure[J]. Physica A: Statistical Mechanics and its Applications, 2017, 477: 149-160.
[6]
Rahman M M, Farahani M A, Wuest T. Multivariate time-series classification of critical events from industrial drying hopper operations: a deep learning approach[J]. Journal of Manufacturing and Materials Processing, 2023, 7(5): 164.
[7]
Gupta A, Gupta H P, Biswas B, et al. A fault-tolerant early classification approach for human activities using multivariate time series[J]. IEEE Transactions on Mobile Computing, 2020, 20(5): 1747-1760.
[8]
Zhang N, Sun S. Multiview unsupervised shapelet learning for multivariate time series clustering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 45(4): 4981-4996.
[9]
Orsenigo C, Vercellis C. Combining discrete SVM and fixed cardinality warping distances for multivariate time series classification[J]. Pattern Recognition, 2010, 43(11): 3787-3794.
[10]
Cheng Z, Yang Y, Wang W, et al. Time2graph: revisiting time series modeling with dynamic Shapelets[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2020: 3617-3624.
[11]
Cho S, Chang W, Lee G, et al. Interpreting internal activation patterns in deep temporal neural networks by finding prototypes[C]// Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. New York: Association for Computing Machinery, 2021: 158-166.
[12]
Wu G, Zhang H, He Y, et al. Learning Kullback-Leibler divergence-based Gaussian model for multivariate time series classification[J]. IEEE Access, 2019, 7: 139580-139591.
[13]
Berndt D J, Clifford J. Using dynamic time warping to find patterns in time series[C]// Proceedings of the 3rd International Conference on Knowledge Discovery and Data Mining. Palo Alto: AAAI, 1994: 359-370.
[14]
Zhao B, Xing H, Wang X, et al. Classification-oriented distributed semantic communication for multivariate time series[J]. IEEE Signal Processing Letters, 2023, 30: 369-373.
[15]
Le X M, Tran M T, Huynh V N. Learning perceptual position-aware shapelets for time series classification[C]// Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Cham: Springer Nature Switzerland, 2022: 53-69.
[16]
李可, 燕晗, 顾杰斐, 等. 基于Shapelets时间序列的多源迁移学习滚动轴承故障诊断方法[J]. 中国机械工程, 2022, 33(24): 2990-2996.
Li Ke, Yan Han, Gu Jiefei, et al. Multi-source transfer learning rolling bearing fault diagnosis method based on Shapelets time series[J]. China Mechanical Engineering, 2022, 33(24): 2990-2996.
[17]
Li G, Yan W, Wu Z. Discovering Shapelets with key points in time series classification[J]. Expert Systems with Applications, 2019, 132: 76-86.
[18]
Hssayni E H, Joudar N E, Ettaouil M. A deep learning framework for time series classification using normal cloud representation and convolutional neural network optimization[J]. Computational Intelligence, 2022, 38(6): 2056-2074.
[19]
Liu C L, Hsaio W H, Tu Y C. Time series classification with multivariate convolutional neural network[J]. IEEE Transactions on Industrial Electronics, 2018, 66(6): 4788-4797.
[20]
Liu M, Ren S, Ma S, et al. Gated transformer networks for multivariate time series classification[J]. ArXiv Preprint ArXiv, 2021: 2103. 14438.
[21]
孙其法. 基于多样化Shapelets的不平衡时间序列分类方法[J]. 信息与电脑(理论版), 2021, 33(18): 45-47.
Sun Qifa. Unbalanced time series classification method based on diversified Shapelets[J]. Information and Computers (Theory), 2021, 33(18): 45-47.
[22]
Ye L, Keogh E. Time series Shapelets: a new primitive for data mining[C]// International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2009: 947-956.
[23]
Liang Z, Wang H. FedST: federated Shapelet transformation for interpretable time series classification[J]. ArXiv Preprint ArXiv, 2023: 2302. 10631.
[24]
Yang J, Jing S, Huang G. Accurate and fast time series classification based on compressed random shapelet forest[J]. Applied Intelligence, 2023, 53(5): 5240-5258.
[25]
Lines J, Davis L M, Hills J, et al. A Shapelet transform for time series classification[C]// Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: Association for Computing Machinery, 2012: 289-297.
[26]
Grabocka J, Schilling N, Wistuba M, et al. Learning time-series shapelets[C]// Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: Association for Computing Machinery, 2014: 392-401.
[27]
Guillaume A, Vrain C, Elloumi W. Random dilated shapelet transform: a new approach for time series shapelets[C]// International Conference on Pattern Recognition and Artificial Intelligence. Cham: Springer International Publishing, 2022: 653-664.
[28]
Chen J, Wan Y. Localized Shapelets selection for interpretable time series classification[J]. Applied Intelligence, 2023, 53(14): 17985-18001.
[29]
Liu H Y, Gao Z Z, Wang Z H, et al. Time series classification with Shapelet and canonical features[J]. Applied Sciences, 2022, 12(17): 8685.
[30]
王威娜, 胡佳利, 任艳. 基于优化Shapelet的时间序列分类方法[J]. 科学技术与工程, 2023, 23(8): 3345-3353.
Wang Weina, Hu Jiali, Ren Yan. Time series classification method based on optimized Shapelet[J]. Science Technology and Engineering, 2023, 23(8): 3345-3353.
[31]
李晨, 万源. 基于优化和两阶段筛选的时间序列Shapelets提取研究[J]. 计算机科学, 2023, 50(2): 146-157.
Li Chen, Wan Yuan. Research on extraction of time series Shapelets based on optimization and two-stage screening[J]. Computer Science, 2023, 50(2): 146-157.
[32]
杨骏, 敬思远, 钟勇. 面向时间序列有序分类的Shapelet抽取算法[J]. 电子科技大学学报, 2023, 52(6): 887-896.
Yang Jun, Jing Siyuan, Zhong Yong. Shapelet extraction algorithm for ordered classification of time series[J]. Journal of University of Electronic Science and Technology, 2023, 52(6): 887-896.
[33]
闫汶和, 李桂玲. 基于Shapelet的时间序列分类研究[J]. 计算机科学, 2019, 46(1): 29-35.
Yan Wenhe, Li Guiling. Research on time series classification based on Shapelet[J]. Computer Science, 2019, 46(1): 29-35.
[34]
Zhang N, Sun S. Multiview unsupervised Shapelet learning for multivariate time series clustering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 45(4): 4981-4996.
[35]
Hills J, Lines J, Baranauskas E, et al. Classification of time series by Shapelet transformation[J]. Data Mining and Knowledge Discovery, 2014, 28: 851-881.
[36]
Shokoohi-Yekta M, Wang J, Keogh E J. On the non-trivial generalization of dynamic time warping to the multi-dimensional case[C]// Proceedings of the 2015 SIAM International Conference on Data Mining. Vancouver, Canada: SIAM, 2015: 289-297.
[37]
Li G, Choi B, Xu J, et al. Shapenet: a Shapelet-neural network approach for multivariate time series classification[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2021: 8375-8383.
[38]
Karlsson I, Papapetrou P, Boström H. Generalized random Shapelet forests[J]. Data Mining and Knowledge Discovery, 2016, 30(5): 1053-1085.
[39]
宋奎勇, 王念滨, 王红滨. 基于 Shapelets 的多变量 DS 证据加权集成分类[J]. 吉林大学学报 (信息科学版), 2021, 39(2): 205-214.
Song Kuiyong, Wang Nianbin, Wang Hongbin. Multivariate DS evidence weighted ensemble classification based on Shapelets[J]. Journal of Jilin University (Information Science Edition), 2021, 39(2): 205-214.
[40]
詹熙, 黎维, 潘志松. Multi-Shapelet: 一种基于Shapelet的多变量时间序列分类方法[J]. 数据采集与处理, 2023, 38(2): 386-400.
Zhan Xi, Li Wei, Pan Zhisong. Multi-Shapelet: a multivariable time series classification method based on Shapelet[J]. Data Acquisition and Processing, 2023, 38(2): 386-400.
[41]
Bagnall A, Dau H A, Lines J, et al. The UEA multivariate time series classification archive, 2018[J]. ArXiv Preprint ArXiv, 2018: 1811.00075.
[42]
Zhang X, Gao Y, Lin J, et al. Tapnet: multivariate time series classification with attentional prototypical network[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2020: 6845-6852.
[43]
Zuo J, Zeitouni K, Taher Y. Smate:semi-supervised spatio-temporal representation learning on multivariate time series[C]//2021 IEEE International Conference on Data Mining (ICDM). Piscataway: IEEE, 2021: 1565-1570.
[44]
Chen R, Yan X, Wang S, et al. DA-Net: dual-attention network for multivariate time series classification[J]. Information Sciences, 2022, 610: 472-487.
[45]
Du M, Wei Y, Zheng X, et al. Multi-feature based network for multivariate time series classification[J]. Information Sciences, 2023, 639: 119009.
2025年第25卷第1期
PDF下载
228
73
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2308252
  • 接收时间:2023-10-23
  • 首发时间:2025-07-29
  • 出版时间:2025-01-08
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2023-10-23
  • 修回日期:2024-10-08
基金
国家自然科学基金(62266046)
吉林省自然科学基金(YDZJ202201ZYTS603)
吉林省教育厅科研项目(JJKH20230281KJ)
作者信息
    吉林化工学院信息与控制工程学院, 吉林 132022
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2308252
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数

Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)

Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏