U-Mixer: An Unet-Mixer Architecture with Stationarity Correction for Time Series Forecasting
Xiang Ma, Xuemei Li, Lexin Fang, Tianlong Zhao, Caiming Zhang
Abstract
Time series forecasting is a crucial task in various domains. Caused by factors such as trends, seasonality, or irregular fluctuations, time series often exhibits non-stationary. It obstructs stable feature propagation through deep layers, disrupts feature distributions, and complicates learning data distribution changes. As a result, many existing models struggle to capture the underlying patterns, leading to degraded forecasting performance. In this study, we tackle the challenge of non-stationarity in time series forecasting with our proposed framework called U-Mixer. By combining Unet and Mixer, U-Mixer effectively captures local temporal dependencies between different patches and channels separately to avoid the influence of distribution variations among channels, and merge low- and high-levels features to obtain comprehensive data representations. The key contribution is a novel stationarity correction method, explicitly restoring data distribution by constraining the difference in stationarity between the data before and after model processing to restore the non-stationarity information, while ensuring the temporal dependencies are preserved. Through extensive experiments on various real-world time series datasets, U-Mixer demonstrates its effectiveness and robustness, and achieves 14.5% and 7.7% improvements over state-of-the-art (SOTA) methods.
BibTeX
@article{Ma_Li_Fang_Zhao_Zhang_2024, title={U-Mixer: An Unet-Mixer Architecture with Stationarity Correction for Time Series Forecasting}, volume={38}, url={https://ojs.aaai.org/index.php/AAAI/article/view/29337}, DOI={10.1609/aaai.v38i13.29337}, abstractNote={Time series forecasting is a crucial task in various domains. Caused by factors such as trends, seasonality, or irregular fluctuations, time series often exhibits non-stationary. It obstructs stable feature propagation through deep layers, disrupts feature distributions, and complicates learning data distribution changes. As a result, many existing models struggle to capture the underlying patterns, leading to degraded forecasting performance. In this study, we tackle the challenge of non-stationarity in time series forecasting with our proposed framework called U-Mixer. By combining Unet and Mixer, U-Mixer effectively captures local temporal dependencies between different patches and channels separately to avoid the influence of distribution variations among channels, and merge low- and high-levels features to obtain comprehensive data representations. The key contribution is a novel stationarity correction method, explicitly restoring data distribution by constraining the difference in stationarity between the data before and after model processing to restore the non-stationarity information, while ensuring the temporal dependencies are preserved. Through extensive experiments on various real-world time series datasets, U-Mixer demonstrates its effectiveness and robustness, and achieves 14.5% and 7.7% improvements over state-of-the-art (SOTA) methods.}, number={13}, journal={Proceedings of the AAAI Conference on Artificial Intelligence}, author={Ma, Xiang and Li, Xuemei and Fang, Lexin and Zhao, Tianlong and Zhang, Caiming}, year={2024}, month={Mar.}, pages={14255-14262} }