基于弧度特征的火箭时序数据相似性评估  

Similarity evaluation of rocket time series data based on radian features

作　　者：曾腾 徐海洲 李林峰[1] 周淦 孟令刚 Zeng Teng;Xu Haizhou;Li Linfeng;Zhou Gan;Meng Linggang(National Computer System Engineering Research Institute of China,Beijing 100083,China;School of Computer Science and Technology,Xidian University,Xi’an 710071,China;School of Computer Science&Technology,Xi’an University of Posts&Telecommunications,Xi’an 710061,China)

机构地区：[1]华北计算机系统工程研究所,北京100083 [2]西安电子科技大学计算机科学与技术学院,陕西西安710071 [3]西安邮电大学计算机学院,陕西西安710061

出　　处：《电子技术应用》2024年第2期71-75,共5页Application of Electronic Technique

摘　　要：评估火箭时序数据的相似性是火箭时序数据分析的主要任务之一。动态时间规整算法是最具代表性的相似性度量算法,但由于其容易发生病态对齐现象,时刻点常被算法错误匹配,导致度量精度难以满足要求。为解决该问题,提出一种基于弧度特征的时序数据相似性评估算法。该算法充分考虑了原始时序特征和弧度特征,并采用时刻邻域信息进行计算,极大地提升了算法对序列局部形状的捕捉能力。将算法用于时间序列分类任务,在9个具有火箭数据类似特征的数据集上与4种相似度度量进行了对比,获得了26.04%以上的分类精度提升,证明了算法的有效性。Evaluating the similarity of rocket time series data is one of the main tasks in rocket time series data analysis.Dynamic time warping(DTW)is the most representative similarity measurement algorithm,but due to its susceptibility to pathological alignment,time points are often mistakenly matched,making it difficult to meet the accuracy requirements in the rocket field.To address this issue,this paper proposes a similarity evaluation algorithm for time series data based on radian features.By fully considering the original temporal features and radian features,and using time neighborhood information for calculation,the algorithm's ability to capture the local shape of the sequence has been greatly improved.The proposed algorithm was applied to time series classification task and achieved a classification accuracy improvement of over 26.04%on nine datasets with similar features of rocket data,which proved its effectiveness.

关 键 词：火箭时序数据 动态时间规整 时间序列相似度 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]