Table 3 Performance (five-fold cross validation) of the predictive models for each serotype when using PCC for feature selection. SVM: Support Vector Machine; Sn: Sensitivity; Sp: Specificity; Acc: Accuracy; MCC: Matthews Correlation Coefficient
Serotype | Feature selection | Classifiers | Bin size | Number of features | Sn | Sp | Acc | MCC |
|---|---|---|---|---|---|---|---|---|
I a | PCC | Random Forest | 9 | 35 | 100.0% | 96.1% | 97.3% | 0.939 |
SVM | 9 | 28 | 100.0% | 99.6% | 99.8% | 0.994 | ||
I b | Random Forest | 3 | 38 | 100.0% | 93.3% | 95.3% | 0.899 | |
SVM | 3 | 30 | 100.0% | 99.3% | 99.5% | 0.988 | ||
III | Random Forest | 6 | 7 | 91.0% | 90.2% | 90.5% | 0.795 | |
SVM | 6 | 7 | 91.0% | 89.7% | 90.2% | 0.789 | ||
V | Random Forest | 9 | 46 | 100.0% | 92.3% | 94.6% | 0.885 | |
SVM | 9 | 43 | 100.0% | 99.6% | 99.8% | 0.994 | ||
VI | Random Forest | 6 | 31 | 97.3% | 91.6% | 93.7% | 0.872 | |
SVM | 7 | 18 | 94.6% | 93.2% | 93.7% | 0.868 |