多元时间序列相似性搜索研究综述

Survey of similarity search for multivariate time series

多元时间序列相似模式挖掘是数据挖掘领域的研究热点,它主要包括特征表示、相似模式度量和相似性搜索3个方面.目前,大部分研究成果主要集中在特征表示和相似模式度量,相似性搜索则成为制约问题突破的关键环节.为此,主要针对多元时间序列的相似性搜索进行综述,归纳了主要的相似模式度量方法,对比了不同相似模式度量下的序列搜索方法,并分析了不同方法的优缺点,以期为进一步研究多元时间序列相似性搜索提供帮助.

Similar patterns mining for multivariate time series is becoming a hot topic in the area of data mining,which consists of three consecutive parts：Feature representation,similarity measure and similarity search.Most researches mainly focus on feature representation and similarity measure,which make similarity search still a tough problem in similar patterns mining.Therefore,the existing similarity measures are summarized,different methods of similarity search for multivariate time series are compared,their merits and demerits are analyzed,and the further research direction is provided.