基于适应成本导数动态时间弯曲度量的异常检测
首发时间:2021-04-29
摘要:本文针对时间序列数据的一般形状,而不是点对点的函数比较,定义了基于适应成本导数的动态时间弯曲度量,通过将多元时间序列按变量纵向排列把每个变量中数值列作为向量看待,运用AC-DDTW度量多元时间序列间的相似性;结合k中值算法对时间序列进行异常检测,模拟实验表明,该算法可以有效识别一般形状的时间序列。
关键词: 适应成本动态时间弯曲 导数 时间序列异常检测
For information in English, please click here
Anomaly detection based on adaptive cost derivative dynamic time warping distance
Abstract:In this paper, for the general shape of time series data, rather than point-to-point function comparison, adaptive cost derivative dynamic time warping distance(AC-DDTW) is defined. By longitudinal arrangement of multivariate time series and treating the numerical column of each variable as a vector, AC-DDTW is used to measure the similarity among multivariate time series. The simulation results show that the algorithm can identify the time series of general shape effectively.
Keywords: Adaptive cost dynamic time warping Derivative Time series anomaly detection
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
基于适应成本导数动态时间弯曲度量的异常检测
评论
全部评论0/1000