Fig. 5: Multi-horizon comparison of seasonal forecasts across networks.
From: Forecasting influenza activity using machine-learned mobility map
For each of the networks (columns), and each borough (rows), the boxplots (median, IQR, and whiskers at 1.5 IQR) show the posterior predictive distribution for peak time forecast at different time horizons (n = 1000). Similar plots for other targets (peak value, onset time) and other seasons are available in Supplementary Notes 2.
