BatchCP: Forecasting Time-Series Data That Have Change Points

TMLR Paper5376 Authors

Abstract: Many methods for time-series forecasting are known in classical statistics, such as autoregression, moving averages, and exponential smoothing. Some novel, recent approaches for time-series forecasting based on deep learning have shown very promising results already. However, time series often have change points, which can degrade the prediction performance substantially. This paper extends existing frameworks by detecting and including those change points. We show that our method, called BatchCP, performs as well as standard frameworks when there are no change points and considerably better when there are change points. More generally, we show that the batch size provides an effective and surprisingly simple way to deal with change points in architectures in modern forecasting models, such as DeepAR, Transformers, and TFTs.

Submission Length: Regular submission (no more than 12 pages of main content)

Assigned Action Editor: ~Philip_K._Chan1

Submission Number: 5376