Fig. 3

Internal 5-fold cross-validation results of NTCP models. A simple model using only baseline ALBI score until next grade as input is shown as a reference (gray dashed line). The AUROC scores are indicated in the legend for each model. (a) ROC curves for the best logistic regression models from each cross-validation fold (colored solid lines) and the average (black solid line). (b) ROC curves for the best random forest models. (c) ROC curves for the best gradient-boosted tree models. (d) Comparison of the average ROC curves for the best logistic regression (PLR, green), random forest (RF, red), and gradient-boosted tree (GBT, blue) models. (e) Comparison of the average ROC curves for the best gradient-boosted tree models (GBT, blue) and the ensemble of the best logistic regression models and the best gradient-boosted tree models (Ensemble, orange)