AdaST: Adaptive Coupling for Spatial-Temporal Forecasting

2026-09-28

AI Summary

This study addresses the spurious dependencies and performance degradation in spatiotemporal prediction caused by overlooking differences in data coupling structures. To this end, we propose a dynamically modulated spatiotemporal modeling framework that introduces a novel decompose-and-recombine paradigm based on heterogeneous experts. Specifically, the framework achieves spatiotemporal feature decoupling through a heterogeneity-aware expert network coupled with a role-alignment module. Subsequently, a correlation-driven adaptive recombiner dynamically adjusts modeling weights according to inherent coupling structures. Extensive experiments demonstrate that the proposed method significantly outperforms existing state-of-the-art baselines, validating the effectiveness of the adaptive strategy in complex scenarios.

Abstract

Spatial-temporal (ST) forecasting underpins many real-world systems such as traffic, climate, and energy networks. While existing methods implicitly assume strong spatiotemporal coupling, we observe that real-world ST data exhibits distinct coupling regimes, ranging from temporal-dominated and spatial-dominated to strongly coupled patterns. This mismatch causes current models to suffer from spurious dependencies and degraded performance when one correlation dominates. To overcome this limitation, we aim to dynamically modulate spatial and temporal modeling based on the data's inherent coupling structure. However, three key challenges exist: unknown coupling structure, heterogeneous coupling dynamics, and suboptimal spatial modeling. We propose AdaST, an adaptive ST forecasting framework that tackles these challenges through a decompose-recompose paradigm. AdaST factorizes inputs into components capturing different coupling patterns using heterogeneity-aware experts. Each component is processed by role-aligned modules, and a correlation-informed adaptive recomposer integrates them for final prediction. Extensive experiments confirm that AdaST significantly outperforms state-of-the-art baselines, validating the necessity of an adaptive approach.

Problem

• spatial-temporal forecasting
• coupling structure
• heterogeneous dynamics
• spurious dependencies
• adaptive modeling

Innovation

• Adaptive Spatio-Temporal Forecasting
• Decompose-Recompose Paradigm
• Heterogeneity-Aware Experts
• Coupling Dynamics
• Correlation-Informed Recomposer