Predicting County-Level COVID-19 Cases using Spatiotemporal Machine Learning: Modeling Human Interactions using Social Media and Cell-Phone Data
Measurements of human interaction through proxies such as social connectedness or movement patterns have proved useful for predictive modeling of COVID-19. In this study, we first compare the power of Facebook’s social connectedness with cell phone-derived human mobility for predicting county-level new cases of COVID-19. Our experiments show that social connectedness is a better proxy for measuring human interactions leading to new infections. Next, we develop a SpatioTemporal autoregressive eXtreme Gradient Boosting (STXGB) model to predict county-level new cases of COVID-19 in the coterminous US. We evaluate the model on five weekly forecast dates between October 24 and November 28, 2020 over one- to four-week prediction horizons. Comparing our predictions with a baseline Ensemble of 32-models currently used by the CDC indicates an average 58% improvement in prediction RMSEs over two- to four-week prediction horizons, pointing to the strong predictive power of our model.
Figure 1
Figure 2
Figure 3
Figure 4
Due to technical limitations, full-text HTML conversion of this manuscript could not be completed. However, the manuscript can be downloaded and accessed as a PDF.
This is a list of supplementary files associated with this preprint. Click to download.
Posted 10 Feb, 2021
Predicting County-Level COVID-19 Cases using Spatiotemporal Machine Learning: Modeling Human Interactions using Social Media and Cell-Phone Data
Posted 10 Feb, 2021
Measurements of human interaction through proxies such as social connectedness or movement patterns have proved useful for predictive modeling of COVID-19. In this study, we first compare the power of Facebook’s social connectedness with cell phone-derived human mobility for predicting county-level new cases of COVID-19. Our experiments show that social connectedness is a better proxy for measuring human interactions leading to new infections. Next, we develop a SpatioTemporal autoregressive eXtreme Gradient Boosting (STXGB) model to predict county-level new cases of COVID-19 in the coterminous US. We evaluate the model on five weekly forecast dates between October 24 and November 28, 2020 over one- to four-week prediction horizons. Comparing our predictions with a baseline Ensemble of 32-models currently used by the CDC indicates an average 58% improvement in prediction RMSEs over two- to four-week prediction horizons, pointing to the strong predictive power of our model.
Figure 1
Figure 2
Figure 3
Figure 4
Due to technical limitations, full-text HTML conversion of this manuscript could not be completed. However, the manuscript can be downloaded and accessed as a PDF.