Using Mobility Data to Understand and Forecast COVID19 Dynamics.

Area of Science:

• Epidemiology
• Computational Biology
• Network Science

Background:

• Pandemic dynamics, including COVID-19, are influenced by evolving human mobility and public policies.
• Accurate forecasting of disease spread requires understanding dynamic human mobility and spatial interaction patterns.

Approach:

• Developed a novel graph-based neural network (GNN) to integrate global aggregated mobility flows.
• Proposed a recurrent message passing GNN model embedding spatio-temporal disease and human mobility dynamics.
• Utilized the GNN for daily state-level new confirmed COVID-19 cases forecasting.

Key Points:

• This research is among the first to apply GNNs for forecasting COVID-19 incidence dynamics.
• The proposed GNN method demonstrates competitive performance against existing forecasting models.
• Leveraging dynamic mobility graphs with GNNs significantly enhances long-term forecasting accuracy compared to baseline methods.

Conclusions:

• Graph neural networks offer a powerful approach for forecasting infectious disease incidence.
• Human mobility data, when integrated through GNNs, significantly improves COVID-19 spread predictions.
• This research contributes to early applications of GNNs in epidemiological forecasting.