An Improved Similarity Comparison Method for Long Time Series

Abstract

In the process of satellite monitoring and control management, a lot of satellite telemetry data were generated and stored at a form of time series in the database. Time series similarity measure is the basis of these technologies, and improving its algorithm can greatly promote the efficiency of anomaly detecting. This paper presents a parallel scheme for fast similarity search based on DTW, called MRDTW, in large satellite time series. By experiment evaluation, we show our approach retained the original accuracy as DTW, and the efficiency of similarity measure has been greatly improved in large time series.