Fig. 3.

Workflow diagram of the machine learning process used to appraise the performance of classification models. The initial data set is split into two subsets: the learning and testing data sets. Afterward, the learning data set undergoes a tenfold cross-validation strategy when training sets are used to select features (“Feature selection” using LASSO logistic regression) and validation sets to test the performance of classification models. (For model building step, we apply LASSO logistic regression over selected features.) Finally, the best model is selected and evaluated on the testing data set (“Evaluation”)