Graph neural ODEs improve long-term epidemic forecasting accuracy
Learning and Forecasting Multiscale Spatiotemporal Epidemic Dynamics with Lagrangian-Eulerian Graph Neural ODEs
Machine Learning
Summary
Predicting how epidemics spread over medium to long periods is hard because the processes are complex and only partially observed. The authors designed a new model called LE-EpiGNODE that uses special graph neural network methods, guided by disease transmission and population movement patterns. This model better predicts epidemic trends by separately modeling infection generation and population flow, which leads to more accurate forecasts, especially over longer times. The approach also offers insights into the underlying epidemic mechanisms, beyond just good predictions.
What this means in practice
- •For public health modelers: Improve medium- to long-term epidemic forecasts using process-aware neural networks that separate infection and mobility effects.
- •For urban planners: Model how population movement affects disease spread to plan interventions or mobility policies in cities during outbreaks.
Authors
Jinyi Duan, Qian Li, Biao Tang
Abstract
Accurate medium- to long-range epidemic forecasting requires extrapolating high-dimensional, heterogeneous and nonstationary dynamics from partial observations. Yet hybrid models often leave implicit which observations inform which epidemic processes and how learned graph interactions modify system state. We introduce LE-EpiGNODE, which replaces generic message passing with process-specific, mechanism-guided neural operators: age-contact information drives Lagrangian infection generation, whereas mobility drives conservative Eulerian population transport within a structure-preserving universal differential equation. We evaluated LE-EpiGNODE across 14 controlled synthetic mechanisms spanning time-modulated, state-dependent, nonlinear and coupled dynamics, three complementary COVID-19 settings, 21 comparator models and five targeted ablations. Forecasting gains tended to become more pronounced at longer horizons; at the longest horizon, LE-EpiGNODE achieved the lowest WMAPE in all three real-world datasets, reducing error by 17.7-40.0% relative to the best baseline. Controlled ablations linked these gains to the Lagrangian-Eulerian decomposition and process-aligned priors, while synthetic ground truth showed that learned operators and induced state fluxes track known generating dynamics. These results support mechanism-organized graph neural dynamics as a route to reliable long-horizon extrapolation with directly testable, process-level mechanistic interpretation beyond trajectory accuracy.