Fig. 3

Calibration curves. The left panel shows calibration curves in the training cohort (Black, ideal line; Blue, extreme gradient boosting (XGBoost); Green, Random Forest (RF); Red: Ensemble model). An ensemble of XGBoost and RF showed the best calibration curve. Thus, the ensemble model was chosen as our final model. The right panel shows the calibration curve of the final model in the test cohort, with the area of pinkish shades indicating 95% confidence interval