Fig. 2: 21-day forecast for each hospital.
From: Forecasting hospital-level COVID-19 admissions using real-time mobility data
Top left: total hospitalizations across all five hospitals. Each other subplot compares the actual to the predicted hospital admissions, based on the model that incorporates mobility and test positivity data as exogenous variables. The vertical dashed line marks the boundary of boundary of initial training data and future data; visualized at a 7-day rolling average; 95% confidence intervals are shown for each forecast. Hyper-parameters of each model were tuned separately to make sure fair comparisons were made between models.
