Table 7.
Comparison of classification accuracies across different feature selection methods in test set
| Classifier | RIMONE | SLO images | ||||
|---|---|---|---|---|---|---|
| wrap-AUC | wrap-LDA | wrap-QDA | wrap-AUC | wrap-LDA | wrap-QDA | |
| Twin SVM | 90.9 % | 86.4 % | 81.8 % | 85.7 % | 78.6 % | 64.3 % |
| Linear SVM | 90.9 % | 88.6 % | 81.8 % | 78.6 % | 71.4 % | 64.3 % |
| Polynomial SVM | 90.9 % | 88.6 % | 81.8 % | 78.6 % | 71.4 % | 64.3 % |
| RBF SVM | 88.6 % | 86.4 % | 97.7 % | 78.6 % | 57.1 % | 78.6 % |
| Sigmoid SVM | 84.1 % | 86.4 % | 86.4 % | 64.3 % | 28.6 % | 28.6 % |
| LDA | 88.6 % | 88.6 % | 79.5 % | 71.4 % | 71.4 % | 64.3 % |
| QDA | 86.4 % | 88.6 % | 90.9 % | 64.3 % | 35.7 % | 74.1 % |