May 26, 2020

Epidemic Model Guided Machine Learning for COVID-19 Forecasts in the United States

Category: Article Summary

Topic: Modeling and Prediction

• [pre-print, not peer reviewed] Zou et al. propose the SuEIR model, which extends the SEIR model by accounting for untested/unreported cases of COVID-19, and train this model using machine learning algorithms. Forecasts from the model predict the peak dates of the outbreak in the US (June 1), New York State (May 10), and California (July 1). Estimated basic reproduction numbers (R0) are 2.5 for the US, 3.6 for New York, and 2.2 for California. Results for all states: https://covid19.uclaml.org. These predictions have been adopted by the CDC for COVID-19 death forecasts.  

Zou et al. (May 25, 2020). Epidemic Model Guided Machine Learning for COVID-19 Forecasts in the United States. Pre-print downloaded May 26 from https://doi.org/10.1101/2020.05.24.20111989