Table 3. Quantitative results of the leave-one-out cross evaluation of the twelve SVM models.
| Model | Exact accuracy | Sliding-window accuracy | Binary accuracy |
|---|---|---|---|
| Simple Model (C = 2.75) | 23.53% | 57.35% | 54.41% |
| Extended Simple Model (C = 2.75) | 23.53% | 77.94% | 72.06% |
| MNI Model (C = 2.55) | 35.29% | 64.71% | 64.71% |
| Extended MNI Model (C = 2.70) | 39.71% | 75.00% | 79.41% |
| HO Subcortical Model (C = 2.30) | 35.29% | 67.65% | 70.59% |
| Extended HO Subcortical Model (C = 2.50) | 35.29% | 73.53% | 79.41% |
| HO Cortical Model (C = 1.00) | 25.00% | 60.29% | 67.65% |
| Extended HO Cortical Model (C = 0.85) | 30.88% | 60.29% | 60.29% |
| Problem-specific Model (C = 2.75) | 41.18% | 66.18% | 67.65% |
| Extended Problem-specific Model (C = 2.55) | 55.88% | 82.35% | 85.29% |
| VLSM Model (C = 2.35) | 25.00% | 52.94% | 60.29% |
| Extended VLSM Model (C = 2.65) | 25.00% | 50.00% | 61.76% |