As for forward selection, the number of attributes was also set to 30, so to keep uniformity<br>throughout this work’s course. The number of rounds performed after the stop criterion is reached<br>was set to 5. This value is higher than the default value proposed by the software because of the<br>intention to avoiding a locally optimal value. This way, the selection of the best features is better<br>assured. This operator is a nested operator, which means it runs a subprocess. In this subprocess,<br>a classifier is used to evaluate performance of the features. This classification task also requires<br>the use of cross-validation, but, in this case, it was decided not to perform leave-one-out cross<br>validation because it was very time consuming. Instead, 8 fold-validation was performed, meaning<br>that 8 groups were created and the classifier ran 8 times, having one group as test set and the<br>remaining 7 as train sets. For the cross-validation operator of the main classifier, leave-one-out<br>cross-validation was still performed.Type your comment