The SuEIR (see the figure below) model is an epidemic model for forecasting the active cases and deaths of COVID-19 by considering the unreported cases. The critical feature of SuEIR is the new compartment called "Unreported". Specifically, we treat the "Exposed" group in SEIR as the individuals that have already been infected but not been tested, who are also capable of infecting others. Some of the people in the "Exposed" group will be tested and transferred to the "Infectious" group (which are reported to the public), while the rest will recover/die and transit to the "Unreported" group that are not reported to the public due to testing capacity, asymptomatic infection, quick recovering, etc.
Illustration of the SuEIR model.
In addition to forecasting the numbers of confirmed cases and deaths, the SuEIR++ model incoporates "Hospitalized" and "ICU" compartments into the SuEIR model to enable the forecasting of the hospitalization resources (e.g., hospitalization and ICU beds). The model is illustrated in the figure below.
Illustration of the SuEIR++ model.
We include all history projections (confirmed cases, deaths, and hospitalizations) in the folder projection_result/, which are updated on a weekly basis. We also visualize our forecasts at county, state, and nation levels in our website covid19.uclaml.org.
For more technical details, please refer to our manuscript.
@article{zou2020epidemic,
title={Epidemic Model Guided Machine Learning for COVID-19 Forecasts in the United States},
author={Zou, Difan and Wang, Lingxiao and Xu, Pan and Chen, Jinghui and Zhang, Weitong and Gu, Quanquan},
journal={medRxiv},
year={2020},
publisher={Cold Spring Harbor Laboratory Press}
}