Probabilistic forecasting of daily COVID-19 admissions using machine learning

Abstract Accepted by: Aris Syntetos Accurate forecasts of daily Coronavirus-2019 (COVID-19) admissions are critical for healthcare planners and decision-makers to better manage scarce resources during around infection peaks. Numerous studies have focused on forecasting COVID-19 at the national or global levels. Localized predictions vital, as they allow resource planning redistribution, but also harder get right. Several possible indicators can be used predict admissions. The inherent variability in necessitates generation evaluation forecast distribution admissions, opposed producing only a point forecast. In this study, we propose quantile regression forest (QRF) model probabilistic local hospital trust (aggregation 3 hospitals), up 7 days ahead, using multitude different predictors. We evaluate accuracy well appropriate measures. provide evidence that QRF outperforms univariate time series methods other more sophisticated benchmarks. Our findings show lagged total positive cases, tests performed, Google grocery Apple driving most salient Finally, highlight areas where further research is needed.