Table 2.
Optimal SVM parameters of different datasets containing different number of samples for model construction using different combination of features selected using filter methods (7 features) and wrapper methods (11 features), respectively.
| Sample (N) | Feature Selection | Accuracy (SD) (%) | Sensitivity (SD) (%) | Specificity (SD) (%) | log2C | Log2γ |
|---|---|---|---|---|---|---|
| 348 | Filter | 88.33 (0.84) | 90.32 (1.46) | 85.86 (1.18) | 32 | 16 |
| Wrapper | 92.73 (0.79) | 95.81 (0.94) | 88.97 (1.96) | 6.2 | 3.1 | |
| 287 | Filter | 85.19 (1.55) | 92.17 (0.87) | 73.97 (3.48) | 4 | 64 |
| Wrapper | 90.56 (1.37) | 95.14 (2.05) | 85.00 (2.34) | 5.9 | 3 | |
| 231 | Filter | 78.73 (1.57) | 91.08 (1.03) | 63.77 (3.28) | 0.0625 | 64 |
| Wrapper | 85.27 (1.57) | 92.34 (2.41) | 76.35 (2.63) | 4.8 | 3 | |
| 188 | Filter | 77.16 (1.16) | 86.55 (1.79) | 64.91 (1.72) | 0.5 | 8 |
| Wrapper | 79.88 (1.34) | 91.42 (1.32) | 76.35 (4.13) | 6 | 3.2 | |