Using Mobility Data to Understand and Forecast COVID19 Dynamics

Abstract

Disease dynamics, human mobility, and public policies co-evolve during a pandemic such as COVID-19. Understanding dynamic human mobility changes and spatial interaction patterns are crucial for understanding and forecasting COVID-19 dynamics. We introduce a novel graph-based neural network(GNN) to incorporate global aggregated mobility flows for a better understanding of the impact of human mobility on COVID-19 dynamics as well as better forecasting of disease dynamics. We propose a recurrent message passing graph neural network that embeds spatio-temporal disease dynamics and human mobility dynamics for daily state-level new confirmed cases forecasting. This work represents one of the early papers on the use of GNNs to forecast COVID-19 incidence dynamics and our methods are competitive to existing methods. We show that the spatial and temporal dynamic mobility graph leveraged by the graph neural network enables better long-term forecasting performance compared to baselines.

Figure 1:

An example of two-hop RMP architecture. Temporal node feature and edge feature vectors are encoded using the feature encoder module. A two-hop RMP module is used to further embed spatio-temporal information to hidden representations. The output module makes the final predictions.

Figure 2:

Pearson correlation between MF and new confirmed cases, together with state level social distancing mandates including emergency declaration (purple), school closure (orange), and stay-at-home (blue) are marked. A boxplot is used to display variation in samples of 53 states each week. The median value is shown along with the median line.

Figure 3:

Heatmap of Pearson correlation matrix of state level time series of new confirmed cases. (3a) training data and (3b) testing data. We can observe that the pattern of correlations between states changed dramatically from training dataset to testing dataset.