Figure 3.

(A) Receiver operator characteristic curves for prediction models. Shown are top five models, including random forest combined and neural network combined (use big data and known inputs), random forest and logistic regression (use only known inputs), and neural network (only big data inputs). (B) Calibration curves (top) and histograms (bottom) for prediction models. Shown are top five models, including random forest combined and neural network combined (use big data and known inputs), random forest and logistic regression (use only known inputs), and neural network (only big data inputs). ROC: Receiver operator characteristic.