Generative diffusion models for spatiotemporal influenza forecasting

2026-04-27

AI Summary

This study addresses the challenge of modeling multimodal uncertainties and abrupt trends in influenza outbreak forecasting by proposing a novel approach based on denoising diffusion probabilistic models. The method encodes flu seasons as spatiotemporal images and employs a conditional generation (inpainting) strategy to produce diverse, high-fidelity future epidemic trajectories from partial observations. Innovatively, it introduces diffusion models to influenza spatiotemporal prediction for the first time and constructs a hybrid training dataset combining real surveillance data with simulated scenarios. Evaluated in the CDC FluSight 2024–2025 challenge, the approach achieves strong performance using a training configuration of 30% real and 70% synthetic data, attaining prediction accuracy comparable to current state-of-the-art ensemble methods.

Abstract

Forecasting infectious disease incidence can provide important information to guide public health planning, yet is difficult because epidemic dynamics are complex. Current mechanistic and statistical approaches often struggle to capture multimodal uncertainty or emergent trends. Influpaint adapts denoising diffusion probabilistic models to epidemic forecasting. By encoding influenza seasons as spatiotemporal images in which pixel intensity represents incidence, Influpaint learns a rich distribution of disease dynamics from a hybrid dataset of surveillance and simulated trajectories. Forecasting is formulated as a conditional generation (inpainting) task from partial observations. We show that Influpaint generates realistic, diverse epidemic trajectories and achieves forecast accuracy that is competitive with leading ensemble methods in retrospective evaluation. In real-time evaluation during the 2023--2025 U.S. CDC FluSight challenges, performance improved substantially across seasons, with highly accurate but somewhat overconfident projections in 2024--2025. The best performance was achieved with a training dataset containing 30% surveillance and 70% simulated trajectories. These results show that diffusion models can capture important spatiotemporal structure in influenza dynamics and provide a flexible framework for probabilistic infectious disease forecasting.

Problem

• influenza forecasting
• spatiotemporal dynamics
• epidemic prediction
• multimodal uncertainty
• infectious disease

Innovation

• diffusion models
• spatiotemporal forecasting
• inpainting
• influenza dynamics
• probabilistic forecasting