Table 2.
Classification accuracies of the actual EEG data.
| Vote threshold | Outside of CV | Within CV | ||||
|---|---|---|---|---|---|---|
| n | SVM | RLDA | n | SVM | RLDA | |
| ≥ 50 | 54 | 77.65 | 79.06 | 37–67 | 69.10 | 68.65 |
| ≥ 60 | 45 | 77.76 | 81.91 | 30–53 | 67.69 | 69.18 |
| ≥ 70 | 39 | 77.53 | 78.92 | 23–41 | 69.13 | 68.20 |
| ≥ 80 | 31 | 75.97 | 79.79 | 15–32 | 67.58 | 68.25 |
| ≥ 90 | 19 | 76.56 | 79.54 | 9–21 | 66.61 | 67.16 |
| Mean (S.D.) | 77.09 (0.71) | 79.84 (1.08) | 68.02 (0.97) | 68.29 (0.66) | ||
Two feature selection approaches (outside of CV and within CV) were tested using two different classifiers (SVM and RLDA, unit: %) with respect to the different vote thresholds (number of selected features).
n* number of selected features, S.D standard deviation, SVM support vector machine, RLDA regularized linear discriminant analysis.