Fig. 3From: Development and validation of ‘Patient Optimizer’ (POP) algorithms for predicting surgical risk with machine learningLength-of-stay: Training data are segmented into 3 classes, to cast predicting length-of-stay as multiclass classification. There is a clear periodicity around whole days. The x-axis is truncated at 200 hours to provide detail in the most interesting range. The trend continues past 200 hours with a steady monotonic decrease in magnitudeBack to article page