Figure 1.

Validation strategy. Flow chart of the validation strategy. First all combinations of feature selection and classification algorithms (4x9) are parametrized in the inner k-fold cross-validation loop based on the training folds of the outer k-fold cross-validation. The best parameters are found by maximizing AUC. Once the parameters are fix the outer k-fold cross-validation loop is run and the average AUC (or similar metric) is recorded.