What question did this study set out to answer?

This research aims to improve time series forecasting by integrating period-aware structures and external factors.

April 4, 2026Open Access

P2D-former for time-series forecasting with period-aware 2D representations and exogenous covariates

HHHu HongChongqing University of Science and Technology GCGuorong ChenChongqing University of Science and Technology WJWang JianChongqing University of Science and Technology

Key Points

This research aims to improve time series forecasting by integrating period-aware structures and external factors.
Developed the P2D-Former model using patch-based transformer architecture.
Incorporated MSUnit and P2D-FFN for modeling periodic structures.
Integrated exogenous variables through an Intersect-Attention mechanism.
Evaluated model performance on multiple benchmark datasets.
P2D-Former achieved higher forecasting accuracy compared to existing methods.
Demonstrated robustness in long-term and short-term forecasting tasks.
Significant improvements were observed in scenarios with high missing values.

Abstract

Time series forecasting finds extensive application across numerous real-world scenarios, yet faces challenges in modelling multi-period variability and exogenous variables. To address this, this paper proposes P2D-Former, a patch-based Transformer model integrating period-aware 2D structures with exogenous information fusion. This approach incorporates MSUnit and P2D-FFN to model complex periodic structures, while explicitly integrating exogenous variable information through an Intersect-Attention mechanism. Experiments demonstrate that P2D-Former achieves superior performance across multiple benchmark datasets for long-term forecasting, short-term forecasting, and high-missing-value imputation tasks, exhibiting significant improvements in both accuracy and robustness compared to existing approaches.

AIに質問

Bookmark

View Full Paper