Repeated Pattern Detection from Time Series

What is it about?

Repeated patterns exist in almost every time series, which characterize many interesting phenomena in reality. Existing methods to detect the repeated pattern are mostly sliding-window based, which is inefficient and limited by the prior knowledge of the target pattern. This paper proposed an efficient and robust method called "mSIMPAD", to detect repeated patterns from time series. The method is coded in Python and is publicly available on Github.

Featured Image

Why is it important?

The proposed method is generic that can detect repeated patterns from multi-variate time series efficiently. It can be adopted as a subroutine in many applications dealing with time series, such as anomaly detection, time series segmentation, time series classification, e.t.c.

Perspectives