Fig. 1

Pipeline applied to train and evaluate the different models used in this study. The input data were divided into ten folds according to a cross-validation (CV) scheme. Feature selection and hyperparameter optimization (HPO) were conducted using a nested CV applied to the training data. The resulting model from this nested CV was then used to make the predictions on the test set. The final performance metrics were calculated based on the average performance of the predictions obtained on the test set