基于相似度的异常检测方法

SIMILARITY BASED TIME SERIES ANALYSE METHOD

提出一种时间序列的相似度计算方法,采用统计学中相关因子的思想,用来衡量两段时间序列数据在视觉上的相似程度。在在各种时间序列数据实时检测中,能快速检测到具有周期性数据的异常并定位异常源,实验证明用于异常检测时效果较好。

Based on the coefficient of time series data,we developed a method to measure the similarity of the data in figure. By using this method,we could detect the anomaly of the time series data with periodicity. In the experiment,the method is proved effectively.