The recall rate and FPR of the models as the features increase. On the left, the features are selected only from empirical features (ID 1-6); On the right, the features are selected only from statistical features (ID 7-14). All of the models were trained using the SMOTE data set and their features increase based on RFE algorithm.