Table 5 Prediction Performance (%): Feature Selection Methods Applied to Training Subsampling of Dataset during Each Validation (The last column lists the mean and standard deviation of the prediction performance over the permutation test (T = 100))
Classifier | Feature Selection | 52-20-split Validation (50 trials) | 12-fold CV (10 trials) | LOOCV | |||
|---|---|---|---|---|---|---|---|
Accuracy | Sensitivity | Specificity | Permutation Test Results | ||||
SVM | SVMRFE | 67.7 | 71.4 | 69.4 | 70.3 | 68.6 | - |
SVM | L1SVM | 72.9 | 76.8 | 76.4 | 78.4 | 74.3 | 49.8 ± 7.4 |
SVM_NL | SVMRFE_NL | 71.6 | 74 | 80.6 | 83.8 | 77.1 | - |
SVM_NL | SVMRW | 61.9 | 68.2 | 70.8 | 67.6 | 74.3 | - |