基于离均差的时间序列相似性度量  

Similarity Measure of Time Series Based on Mean Deviations

作　　者：曹洋洋[1] 林意[1] 

机构地区：[1]江南大学数字媒体学院,江苏无锡214122

出　　处：《计算机与现代化》2017年第5期76-82,87,共8页Computer and Modernization

摘　　要：时间序列的相似性度量是数据挖掘领域研究的一个热点,高维多元时间序列数据一般含有大量的噪声不利于相似性的比较。针对现有的时间序列度量方法存在的问题,在改进的时间序列自底向上融合算法的基础上,提出一种新的基于离均差的时间序列相似性度量的夹角余弦算法(Angle Cosine Metric Similarity,ACMS)。ACMS算法将时间序列等价为一个多维度的向量,充分考虑2个向量的方向和大小特征,增强振幅变化的鲁棒性,减少人为干扰,对数据挖掘中的聚类和预测具有帮助作用。The similarity measure of time series is a hot spot in the field of data mining. High dimensional multivariate time series data generally contain a large amount of noise which is not conducive to the comparison of similarity. In view of the problems in the existing time series measure method, the paper presents a new angle cosine algorithm based on the piecewise linear representation of time series shortly called ACMS. ACMS algorithm will be equivalent to a multi-dimensional vector of time series, the direction and size characteristics of the two vectors are fully considered, which enhances the robustness of time and amplitude variation, reduces human disturbance, and is helpful to the clustering and prediction in data mining.

关 键 词：时间序列 向量 夹角余弦 相似性度量 离均差 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]