Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhang, Xingyu, Du, Hanyun, Song, Zeen, Zhao, Siyu, Zheng, Changwen, Qiang, Wenwen
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2505.16308
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915795175473152
author Zhang, Xingyu
Du, Hanyun
Song, Zeen
Zhao, Siyu
Zheng, Changwen
Qiang, Wenwen
author_facet Zhang, Xingyu
Du, Hanyun
Song, Zeen
Zhao, Siyu
Zheng, Changwen
Qiang, Wenwen
contents Most existing multivariate time series forecasting methods adopt an all-to-all paradigm that feeds all variable histories into a unified model to predict their future values without distinguishing their individual roles. However, this undifferentiated paradigm makes it difficult to identify variable-specific causal influences and often entangles causally relevant information with spurious correlations. To address this limitation, we propose an all-to-one forecasting paradigm that predicts each target variable separately. Specifically, we first construct a Structural Causal Model from observational data and then, for each target variable, we partition the historical sequence into four subsegments according to the inferred causal structure: endogenous, direct causal, collider causal, and spurious correlation. Furthermore, we propose the Causal Decomposition Transformer (CDT), which integrates a dynamic causal adapter to learn causal structures initialized by the inferred graph, enabling correction of imperfect causal discovery during training. Furthermore, motivated by causal theory, we apply a projection-based output constraint to mitigate collider induced bias and improve robustness. Extensive experiments on multiple benchmark datasets demonstrate the effectiveness of the CDT.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16308
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Beyond All-to-All: Causal-Aligned Transformer with Dynamic Structure Learning for Multivariate Time Series Forecasting
Zhang, Xingyu
Du, Hanyun
Song, Zeen
Zhao, Siyu
Zheng, Changwen
Qiang, Wenwen
Machine Learning
Most existing multivariate time series forecasting methods adopt an all-to-all paradigm that feeds all variable histories into a unified model to predict their future values without distinguishing their individual roles. However, this undifferentiated paradigm makes it difficult to identify variable-specific causal influences and often entangles causally relevant information with spurious correlations. To address this limitation, we propose an all-to-one forecasting paradigm that predicts each target variable separately. Specifically, we first construct a Structural Causal Model from observational data and then, for each target variable, we partition the historical sequence into four subsegments according to the inferred causal structure: endogenous, direct causal, collider causal, and spurious correlation. Furthermore, we propose the Causal Decomposition Transformer (CDT), which integrates a dynamic causal adapter to learn causal structures initialized by the inferred graph, enabling correction of imperfect causal discovery during training. Furthermore, motivated by causal theory, we apply a projection-based output constraint to mitigate collider induced bias and improve robustness. Extensive experiments on multiple benchmark datasets demonstrate the effectiveness of the CDT.
title Beyond All-to-All: Causal-Aligned Transformer with Dynamic Structure Learning for Multivariate Time Series Forecasting
topic Machine Learning
url https://arxiv.org/abs/2505.16308