Fig. 2: Feature importance and cumulative AUC for random forest.

The cumulative AUC values represent the average of 10 independent repetitions of model training and performance evaluation. A 100-fold cross-validation strategy was used to ensure robust performance estimation.