Navigating Concept Drift and Temporal Shift: Distribution Shift Generalized Time-Series Forecasting

ICLR 2025 Conference Submission946 Authors

Keywords: Time-Series Forecasting, Distribution Shift Generalization

Abstract:

Time-series forecasting finds broad applications in real-world scenarios. Due to the dynamic nature of time series data, it is crucial for time-series forecasting models to produce robust predictions under potential distribution shifts. In this paper, we initially identify two types of distribution shifts in time series: concept drift and temporal shift. We acknowledge that while existing studies primarily focus on addressing temporal shift issues in time series, designing proper concept drift methods for time series data received comparatively less attention.

Motivated by the need to mitigate potential concept drift issues in time-series forecasting, this work proposes a novel soft attention mechanism that effectively leverages and ensemble information from the horizon time series. Furthermore, recognizing that both concept drift and temporal shift could occur concurrently in time-series forecasting scenarios while an integrated solution remains missing, this paper introduces ShifTS, a model-agnostic framework seamlessly addressing both concept drift and temporal shift issues in time-series forecasting. Extensive experiments demonstrate the efficacy of ShifTS in consistently enhancing the forecasting accuracy of agnostic models across multiple datasets, and consistently outperforming existing concept drift, temporal shift, and combined baselines.

Primary Area: learning on time series and dynamical systems

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 946