Pandemic velocity: Forecasting COVID-19 in the US with a machine learning & Bayesian time series compartmental model

Abstract

Predictions of COVID-19 case growth and mortality are critical to the decisions of political leaders, businesses, and individuals grappling with the pandemic. This predictive task is challenging due to the novelty of the virus, limited data, and dynamic political and societal responses. We embed a Bayesian time series model and a random forest algorithm within an epidemiological compartmental model for empirically grounded COVID-19 predictions. The Bayesian case model fits a location-specific curve to the velocity (first derivative) of the log transformed cumulative case count, borrowing strength across geographic locations and incorporating prior information to obtain a posterior distribution for case trajectories. The compartmental model uses this distribution and predicts deaths using a random forest algorithm trained on COVID-19 data and population-level characteristics, yielding daily projections and interval estimates for cases and deaths in U.S. states. We evaluated the model by training it on progressively longer periods of the pandemic and computing its predictive accuracy over 21-day forecasts. The substantial variation in predicted trajectories and associated uncertainty between states is illustrated by comparing three unique locations: New York, Colorado, and West Virginia. The sophistication and accuracy of this COVID-19 model offer reliable predictions and uncertainty estimates for the current trajectory of the pandemic in the U.S. and provide a platform for future predictions as shifting political and societal responses alter its course.

Leveraging artificial intelligence for pandemic preparedness and response: a scoping review to identify key use cases

“…Early in the pandemic, when data were limited, ML was used to augment traditional modeling approaches. Four studies used neural networks 8 – 11 and one used a random forest algorithm 12 to provide data-driven estimates of parameters for compartmental or statistical models. Two studies compared the performance of neural network-augmented models with traditional Susceptible-(Exposed)-Infected-Recovered models and showed that the augmented models provided better approximations of the true epidemic curve resulting in more accurate forecasts 8 , 10 .…”

Longitudinal COVID-19 Surveillance and Characterization in the Workplace with Public Health and Diagnostic Endpoints

“…A COVID-19 infection model that combines traditional susceptible, exposed, infected, and recovered (SEIR) compartmental models with Bayesian time series and modern machine learning was used for model analysis ( 34 ). This model fuses three methods to provide accurate predictions of case counts and hospitalizations.…”

Impact estimation on COVID-19 infections following school reopening in September 2020 in Italy

“…In particular, a stochastic model of the type S-ARIMA–REG has been employed (to the best of the authors’ knowledge, this is the first time such a method has been applied to address research questions similar to that investigated in this paper). Time dependent data have been extensively employed in epidemiology 24 25 26 27 28 29 37 and modeled by a variety of stochastic models, including S-ARIMA. Introduced in 1970 by Box and Jenkins 33 , these type of models have been successfully used in many fields of research, including epidemiology, see e.g.…”