时间序列数据挖掘中的动态时间弯曲研究综述  

Review on dynamic time warping in time series data mining

作　　者：李海林[1] 梁叶 王少春[1] LI Hai-lin;LIANG Ye;WANG Shao-chun(College of Business Administration,Huaqiao University,Quanzhou 362021,China)

机构地区：[1]华侨大学工商管理学院,福建泉州362021

出　　处：《控制与决策》2018年第8期1345-1353,共9页Control and Decision

基　　金：国家自然科学基金项目(71771094;61300139);福建省社会科学规划项目(FJ2016B076);福建省新世纪优秀人才支持计划项目(Z1625112)

摘　　要：动态时间弯曲是一种重要的相似性度量方法,对时间序列数据挖掘的性能起着至为关键的作用,对其进行全面和深入的探索具有十分重要的理论意义和实际应用价值.首先简述动态时间弯曲算法的基本步骤,并分析其优点和存在的不足;然后,从动态时间弯曲度量效率的改进研究、度量效果的提升措施以及其在各个行业的应用研究等进行相关综述;最后,给出动态时间弯曲的进一步研究方向.通过对动态时间弯曲方法相关综述及分析,能为相似性度量、聚类和分类等时间序列数据挖掘技术提供必要的文献资料和理论基础.As a significant similarity measure, dynamic time warping has critical effect on the performance of time series data mining. The study of dynamic time warping has significant theoretical and practical importance. Firstly,the basic step of the dynamic time warping algorithm is introduced, and the advantages and the shortages are analyzed.Then, dynamic time warping at the aspects of the related improved research of measure efficiency, promotion ways of measure quality and its application in various industries is reviewed. Furthermore, relative some research hotspots are summarized, and several directions for the future research is given. The review and analysis for the related methods of dynamic time warping can provide necessary document literature and theory basis for time series data mining including similarity, clustering and classification.

关 键 词：时间序列 动态时间弯曲 相似性度量 数据挖掘 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]