Fig. 1From: Stable feature selection based on the ensemble L 1 -norm support vector machine for biomarker discoveryBackward feature elimination for optimal subset. Backward feature elimination is performed based on our ranking criteria. SVM with an RBF kernel is used as a classifier for calculating the cross-validation score. The X- and Y-axes denote the size of the feature subset and the 10-fold cross-validation AUC score, respectively. The red circle indicates the number of features with the highest AUC score in our experiment, which is 177 features with AUC = 0.996Back to article page