Figure 1.

Flow-chart of the proposed model. The performance of the models was assessed by using the hold-out method where the dataset was split into two randomly exclusive sets (80% training and 20% test sets). On training sample of hold-out sampling, we have evaluated the important of features by means three different techniques and identified the best features subset by compering the performances of three different classifiers on 100 10-fold cross-validation rounds. Then, we have tested the best features subset on the test set of hold-out sampling.