2.3.5. Model Validation

We tested the predictive performance of models using the receiver operating characteristics, for which we analysed the Area Under Curve (AUC) [66], and the minimum difference between training and testing AUC data (AUCdiff) [67]. These model evaluation statistics range between 0 and 1. Excellent model performances are expressed respectively by AUC values close to 1 and AUCdiff close to 0.
