Table 3.
Comparison of performance of feature selection methods with SVM leave-one-out validation of EEG2.
| No feature selection method | PCA | Wavelet transform | SWLDA | Theta band features | |
|---|---|---|---|---|---|
| EEG2 | 78% | 62% | 67% | 58% | 93% |
As it is seen, whole theta band features give significantly higher accuracy than other methods. The maximum accuracy is in bold.