Fig. 4

Performance of the EL model. (a) Mean ROC curve (with standard deviation shown as shaded area) derived from 1000 bootstrap iterations on the training set (AUC = 0.71 ± 0.01). (b) ROC curve on the independent test set (AUC = 0.69). (c) Confusion matrix on the test set. (d) Performance metrics on the test set: log loss = 0.66, recall = 0.69, accuracy = 0.64, precision = 0.40, F1-score = 0.50, ROC-AUC = 0.69