Table 1.
Metric | Kernel |
||||
---|---|---|---|---|---|
Linear | Polynomial (d = 2) | Polynomial (d = 3) | RBF | Sigmoid | |
Sensitivity | 0.80 ± 0.05 | 0.80 ± 0.05 | 0.82 ± 0.03 | 0.60 ± 0.05 | 0.00 ± 0.00 |
Specificity | 0.91 ± 0.01 | 0.96 ± 0.01 | 0.96 ± 0.01 | 0.99 ± 0.02 | 0.84 ± 0.03 |
Sensitivity = TP/(TP + FN); Specificity = TN/(TN + FP),
where TP, TN, FP, FN correspond to true positive, true negative, false positive and false negative, respectively. We performed three independent runs of 10-fold cross validation on the training collection and reported the average sensitivity/specificity and the standard deviation. The kernel with the best performance in both sensitivity and specificity is highlighted in bold. This is also the kernel we used throughout our analyses.