Table 2. SVM tuning results.
Dataset | Cost | Full dataset (MCR) | Trimmed dataset (MCR) |
Gaucher | 0.1 | 0.204 | 0.05 |
1 | 0.204 | 0.052 | |
5 | 0.204 | 0.052 | |
10 | 0.204 | 0.052 | |
CRC | 0.1 | 0.102 | 0.112 |
1 | 0.105 | 0.139 | |
5 | 0.105 | 0.156 | |
10 | 0.105 | 0.156 | |
Lung | 0.1 | 0.182 | 0.168 |
1 | 0.183 | 0.192 | |
5 | 0.183 | 0.207 | |
10 | 0.183 | 0.208 | |
Ovarian | 0.1 | 0.266 | 0.15 |
1 | 0.266 | 0.165 | |
5 | 0.266 | 0.166 | |
10 | 0.266 | 0.166 |
The performance summary (MCR = Misclassification rate) of a SVM-based classifier for both the full dataset (“full”) and the trimmed dataset (“trimmed”) that underwent variable selection using a univariate moderated t-statistic. These are mean values based on 1000 bootstrap samples for each dataset except for the OC data which used 200 bootstrap samples.