Metapopulation Graph Neural Networks: Deep Metapopulation Epidemic Modeling with Human Mobility

TL;DR

This paper introduces MepoGNN, a hybrid graph neural network model that combines epidemiological knowledge with deep learning to improve multi-region epidemic forecasting and interpretability.

Contribution

It presents a novel end-to-end hybrid model integrating GNNs with the Metapopulation SIR model, explicitly learning epidemiological parameters and propagation graphs from data.

Findings

Outperforms existing models significantly in epidemic prediction accuracy.

Learns interpretable epidemiological parameters and propagation graphs.

Effectively uses generated mobility data when real mobility data is unavailable.

Abstract

Epidemic prediction is a fundamental task for epidemic control and prevention. Many mechanistic models and deep learning models are built for this task. However, most mechanistic models have difficulty estimating the time/region-varying epidemiological parameters, while most deep learning models lack the guidance of epidemiological domain knowledge and interpretability of prediction results. In this study, we propose a novel hybrid model called MepoGNN for multi-step multi-region epidemic forecasting by incorporating Graph Neural Networks (GNNs) and graph learning mechanisms into Metapopulation SIR model. Our model can not only predict the number of confirmed cases but also explicitly learn the epidemiological parameters and the underlying epidemic propagation graph from heterogeneous data in an end-to-end manner. The multi-source epidemic-related data and mobility data of Japan are collected and processed to form the dataset for experiments. The experimental results demonstrate our model outperforms the existing mechanistic models and deep learning models by a large margin. Furthermore, the analysis on the learned parameters illustrate the high reliability and interpretability of our model and helps better understanding of epidemic spread. In addition, a mobility generation method is presented to address the issue of unavailable mobility data, and the experimental results demonstrate effectiveness of the generated mobility data as an input to our model.