Retracted: Time Series Change Point Detection with Contrastive Learning

Abstract

The change Point Detection (CPD) method identifies the time series method related to the trend and attributes the change of time series data to describe the fundamental behavior of the system. For example, detecting changes and anomalies related to web service usage, application usage or human behavior, and system state can provide valuable insights for downstream modeling tasks. We propose comparative learning based on a self-monitoring time series change point detection method. Compared with traditional methods, we can apply them to a broader range. We prove that our method is superior to the state-of-the-art CPD method through experiments on three different and widely used time-series data sets.