Towards the efficacy of federated prediction for epidemics on networks

TL;DR

This paper tackles privacy concerns in epidemic forecasting by proposing a general privacy-preserving federated learning (FL) framework for node-level prediction on networks, implemented via fed-LSTM and fed-STGAT models. It systematically evaluates how model choice, aggregation strategy, graph partitioning, and data quality affect FL performance using a real-world airline network, introducing the efficacy energy metric to measure robustness across varying client configurations. STGAT captures spatio-temporal dependencies better in complex dynamics, while LSTM excels in simpler patterns, with FedProx often yielding more stable performance under heterogeneity. The work provides practical guidelines for deploying FL in epidemic management and suggests broader applications to collective dynamics beyond epidemiology.

Abstract

Epidemic prediction is of practical significance in public health, enabling early intervention, resource allocation, and strategic planning. However, privacy concerns often hinder the sharing of health data among institutions, limiting the development of accurate prediction models. In this paper, we develop a general privacy-preserving framework for node-level epidemic prediction on networks based on federated learning (FL). We frame the spatio-temporal spread of epidemics across multiple data-isolated subnetworks, where each node state represents the aggregate epidemic severity within a community. Then, both the pure temporal LSTM model and the spatio-temporal model i.e., Spatio-Temporal Graph Attention Network (STGAT) are proposed to address the federated epidemic prediction. Extensive experiments are conducted on various epidemic processes using a practical airline network, offering a comprehensive assessment of FL efficacy under diverse scenarios. By introducing the efficacy energy metric to measure system robustness under various client configurations, we systematically explore key factors influencing FL performance, including client numbers, aggregation strategies, graph partitioning, missing infectious reports. Numerical results manifest that STGAT excels in capturing spatio-temporal dependencies in dynamic processes whereas LSTM performs well in simpler pattern. Moreover, our findings highlight the importance of balancing feature consistency and volume uniformity among clients, as well as the prediction dilemma between information richness and intrinsic stochasticity of dynamic processes. This study offers practical insights into the efficacy of FL scenario in epidemic management, demonstrates the potential of FL to address broader collective dynamics.

Figures (11)

• Figure 1: Illustration of an SIR spreading on a network. The left subfigure shows a snapshot at $t=5$, where the blue nodes represent the susceptible, the orange represents the infected, and the greed for the recovered. The upper right subfigure exemplifies temporal node states, and the bottom right subfigure present the prevalence of this SIR spreading.
• Figure 2: Illustration of the proposed federated learning framework designed to epidemic prediction. A central server coordinates the training process, while each client performs effective local training based on its own temporal node states and subnetwork topology.
• Figure 3: Architecture of spatio-temporal graph attention networks (STGAT).
• Figure 4: Illustrations of the fraction of infected nodes as a function of time for various epidemic processes, including SIS, SIR, SEIR, SIRVS, nmSIS, SIStv, and SIRS.
• Figure 5: Comparison of centralized learning and federated learning for an exampled nmSIS process under 4 or 8 clients by the accuracy metric $\alpha=Acc$. The bar represents the performance under solo learning, Fedavg learning and Fedprox learning. The horizontal lines indicate the mean metric under centralized learning, Fedavg learning and Fedprox learning.