Table 3.
Predictive performance indicators of the four kernel functions
| Evaluation index | Training set | Test set | ||||||
|---|---|---|---|---|---|---|---|---|
| Linear-SVM | Polynomial-SVM | Sigmoid-SVM | RBF-SVM | Linear-SVM | Polynomial-SVM | Sigmoid-SVM | RBF-SVM | |
| Precision (%) | 69.89 | 78.07 | 79.37 | 84.21 | 86.36 | 87.50 | 80.77 | 88.24 |
| Recall (%) | 50.78 | 69.53 | 78.74 | 88.19 | 51.35 | 75.68 | 56.76 | 81.08 |
| Accuracy (%) | 83.92 | 88.69 | 90.81 | 93.82 | 85.11 | 90.78 | 85.11 | 92.20 |
| F1 index | 0.59 | 0.74 | 0.79 | 0.86 | 0.64 | 0.81 | 0.67 | 0.85 |
| AUC | 0.722 | 0.819 | 0.866 | 0.918 | 0.742 | 0.858 | 0.759 | 0.885 |
AUC area under the ROC curve, SVM support vector machine