The results of multiple-feature selection based on RFE. The values on the left/right of each table cell are the results of three-feature/four-feature selection. The performance metrics of our models were calculated with four types of sampling, including ordinary training set (noted as ‘NonSampling’), SMOTE training set (’SMOTE’), under-sampling training set (’UnderSample’), and oversampling training set (’OverSample’). Selected features depend on the used classifiers as well as the sampling types. Detailed analyses are made in Section 4.2.2.