单变量序列数据分类方法综述  

Survey of univariate sequence data classification methods

作　　者：张晔 侯毅 欧阳克威 周石琳[1] ZHANG Ye;HOU Yi;OUYANG Kewei;ZHOU Shilin(College of Electronic Science and Technology,National University of Defense Technology,Changsha 410073,China)

机构地区：[1]国防科技大学电子科学学院,湖南长沙410073

出　　处：《系统工程与电子技术》2023年第2期313-335,共23页Systems Engineering and Electronics

基　　金：国家自然科学基金(61903373)资助课题。

摘　　要：单变量序列数据分类涉及现实世界的诸多应用领域,具有重要的研究意义与应用价值。目前,单变量序列数据分类领域的发展处于深度学习逐渐取代传统方法的关键时期,但相关的归纳综述仍然很少。为了促进未来研究,本文对单变量序列数据分类方法进行了全面的总结,根据提取分类信息的不同,将现有分类方法分为基于形状信息、基于频率信息、基于上下文信息以及基于信息融合4种类别。此外,本文依托公开数据集对典型分类方法进行了对比与分析,并对未来研究方向进行了展望。Univariate sequence data classification has wide applications in the real world.Therefore,it has important research significance and application value.At present,due to deep learning is gradually replacing traditional methods,it is a critical developing period for univariate sequence data classification.However,there are few systematic reviews.To stimulate future research,this paper presents a comprehensive review of the univariate sequence data classification methods.These methods are divided into four categories:shape information based methods,frequency information based methods,context information based methods and information fusion based methods,according to different basis of classification.Besides,this paper makes a comparative analysis of typical classification methods based on open data sets,and prospects for future research directions.

关 键 词：单变量序列数据分类 形状信息 频率信息 上下文信息 信息融合 深度学习 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]