Fig. 6: Leave-one-out cross validation performance.
From: Forecasting influenza activity using machine-learned mobility map
a For each of the boroughs, the ground truth is shown in solid line along with the 90% calibrated envelope when the borough’s data were left out of the calibration process. b Boxplots (median, IQR, and whiskers at 1.5 IQR) of the predicted peak time and total ED visits (episize) for the left-out borough (n = 1000), with the ground truth shown by blue dots.
