4.2.4. Implementation

All aforementioned feature selection methods were implemented in Python using the scikit-learn library [33]. The Recursive Feature Elimination is implemented using a Support Vector Classification and uses the accuracy as its performance rating. We specify the average accuracy in percent, with an accuracy of 100 % indicating that all input samples could be correctly categorized. As the collected datasets are rather small, only a 2-fold

cross-validation is conducted. The 2-fold validation is stratified to ensure that the class distributions between the test and training sets remain comparable to the full set of data.
