Beyond Time Series: Spatial Reasoning for Epidemic Forecasting via Multimodal Learning

2026-06-20

AI Summary

This study addresses the limitation of traditional epidemic forecasting models that treat administrative regions as atomic units, thereby neglecting the influence of sub-regional spatial structure on disease transmission. To overcome this, the authors propose a structure-aware multimodal spatiotemporal prediction framework that integrates regional time series with multi-source heterogeneous spatial signals. The model employs an attention mechanism to enable dynamic spatiotemporal joint reasoning, allowing spatial context to adaptively modulate temporal representations. Notably, it introduces structure-aware spatial modeling for the first time, effectively handling heterogeneous data with mismatched resolution and structural properties, while supporting interpretable analysis of when and where spatial signals contribute to predictions. Evaluated on real-time forecasting tasks for COVID-19, influenza, and influenza-like illnesses, the model consistently outperforms state-of-the-art baselines in multivariate time series, multimodal learning, and epidemiological modeling, while maintaining strong probabilistic forecasting performance.

Abstract

Epidemic forecasting models typically rely on surveillance data reported over administrative regions, treating them as atomic units, thereby obscuring sub-regional spatial structure that shapes disease dynamics. We introduce a spatially structured multimodal epidemic forecasting setting that integrates region-level temporal surveillance data with spatially localized auxiliary signals that are misaligned in resolution and structure, reflecting realistic public health reporting constraints. Building on this formulation, we propose M-SPICE (Multimodal SPatIal Context for Epidemic Forecasting), a structure-aware spatiotemporal forecasting framework that performs joint reasoning over temporal disease dynamics and spatial context via attention-based multimodal fusion, allowing spatial signals to selectively condition temporal representations across forecast horizons. We evaluate our approach on real-world COVID-19, influenza, and influenza-like illness (ILI) forecasting tasks under realistic real-time evaluation protocols. Across all forecasting settings, our method consistently outperforms state-of-the-art multivariate time-series, multimodal, and epidemiological forecasting baselines while maintaining strong probabilistic forecasting performance. Finally, interpretability analyses reveal when, where, and how spatial signals are leveraged, highlighting settings in which purely temporal, region-aggregated models are most likely to fail.

Problem

• epidemic forecasting
• spatial reasoning
• multimodal learning
• spatiotemporal modeling
• disease dynamics

Innovation

• spatial reasoning
• multimodal learning
• epidemic forecasting
• attention-based fusion
• spatiotemporal modeling