Table 7.
Method | Balance method | CV | SE | SP | Gm | ACC(%) Multiloop |
---|---|---|---|---|---|---|
Triplet-SVM (SVM) | – | – | 88.40 | 83.50 | 0.859 | N/A |
MiPred (random forest) | – | – | 84.34 | 93.56 | 0.888 | N/A |
microPred (SVM) | SMOTE | outer 5cv | 90.50 | 66.43 | 0.775 | 54.23 |
mirExplorer (AdaBoost) | SMOTE + undersampling | outer 10cv | 94.32 | 97.11 | 0.957 | 92.68 |
Our method (ensemble) | Modified-SMOTEBagging | 10x5cv | 95.11 | 97.91 | 0.965 | 97.25 |
SE, SP and ACC represent sensitivity, specificity and accuracy, respectively.