Table 1. Validation and testing accuracies of each electrode pair for three task phases for various analysis techniques using IHAR.
| ML algorithm | AF3-AF4 | F7-F8 | F3-F4 | FC5-FC6 | T7-T8 | P7-P8 | O1-O2 | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| VA | TA | VA | TA | VA | TA | VA | TA | VA | TA | VA | TA | VA | TA | |
| LDA | 86.7 ± 1.9 | 70.5 ± 2.2 | 91.8 ± 1.4 | 83.5 ± 1.4 | 67.6 ± 1.5 | 50.1 ± 2.4 | 94.3 ± 1.4 | 87.4 ± 1.2 | 82.0 ± 1.6 | 69.1 ± 2.2 | 80.2 ± 1.3 | 65.0 ± 1.5 | 65.2 ± 2.2 | 47.1 ± 2.4 |
| 85.0 ± 1.4 | 76.1 ± 1.3 | 91.2 ± 1.3 | 85.3 ± 1.8 | 62.9 ± 1.8 | 50.8 ± 2.5 | 90.2 ± 1.5 | 78.0 ± 1.7 | 77.2 ± 2.3 | 55.0 ± 2.2 | 79.3 ± 2.2 | 75.8 ± 3.2 | 62.3 ± 2.4 | 51.1 ± 1.7 | |
| 86.9 ± 1.6 | 73.6 ± 1.8 | 92.1 ± 1.1 | 74.6 ± 1.7 | 67.3 ± 1.6 | 60.0 ± 1.3 | 93.4 ± 0.8 | 89.1 ± 0.8 | 78.9 ± 1.9 | 59.8 ± 1.4 | 79.8 ± 1.9 | 77.8 ± 2.0 | 68.6 ± 2.6 | 59.3 ± 2.2 | |
| QDA | 86.0 ± 1.2 | 73.3 ± 1.5 | 94.9 ± 1.0 | 88.3 ± 2.1 | 60.2 ± 2.7 | 57.1 ± 2.3 | 91.8 ± 0.8 | 94.7 ± 1.1 | 79.0 ± 1.6 | 80.6 ± 2.1 | 78.8 ± 1.7 | 73.3 ± 2.6 | 59.8 ± 1.3 | 55.3 ± 3.2 |
| 81.4 ± 2.1 | 90.1 ± 0.5 | 86.1 ± 1.7 | 83.5 ± 1.2 | 50.8 ± 3.1 | 62.8 ± 0.8 | 88.5 ± 1.7 | 82.3 ± 0.8 | 75.8 ± 1.5 | 76.0 ± 0.8 | 75.4 ± 1.7 | 72.5 ± 1.1 | 55.8 ± 3.1 | 51.0 ± 1.4 | |
| 85.1 ± 1.8 | 84.0 ± 1.4 | 94.3 ± 1.3 | 81.5 ± 0.5 | 55.0 ± 2.7 | 55.3 ± 2.1 | 94.9 ± 1.1 | 92.0 ± 0.1 | 76.5 ± 2.0 | 68.5 ± 0.5 | 78.1 ± 1.7 | 76.8 ± 0.9 | 63.1 ± 1.7 | 59.3 ± 0.8 | |
| Linear SVM | 89.1 ± 1.4 | 74.6 ± 1.8 | 93.1 ± 1.1 | 89.1 ± 2.3 | 70.2 ± 2.3 | 47.8 ± 2.3 | 95.1 ± 1.1 | 86.3 ± 2.2 | 84.0 ± 1.1 | 68.8 ± 2.6 | 83.1 ± 2.3 | 61.3 ± 2.3 | 70.6 ± 2.5 | 43.5 ± 2.8 |
| 86.5 ± 1.2 | 73.6 ± 1.5 | 92.7 ± 1.2 | 83.1 ± 1.6 | 65.8 ± 2.1 | 52.5 ± 2.5 | 90.9 ± 0.5 | 79.3 ± 1.6 | 80.6 ± 1.1 | 57.1 ± 1.7 | 82.0 ± 1.8 | 73.3 ± 2.6 | 64.6 ± 2.1 | 47.5 ± 2.7 | |
| 88.7 ± 1.3 | 75.0 ± 2.3 | 93.8 ± 1.4 | 82.8 ± 1.9 | 69.0 ± 3.1 | 61.3 ± 1.3 | 94.1 ± 1.0 | 90.0 ± 0.1 | 80.6 ± 1.3 | 61.6 ± 1.3 | 81.6 ± 2.1 | 76.1 ± 2.3 | 70.9 ± 2.4 | 57.3 ± 1.7 | |
| Quadratic SVM | 93.1 ± 1.3 | 80.5 ± 1.2 | 95.3 ± 1.5 | 85.3 ± 2.6 | 79.2 ± 1.5 | 57.0 ± 2.3 | 96.6 ± 0.1 | 83.1 ± 1.1 | 89.1 ± 0.1 | 77.3 ± 2.1 | 89.7 ± 1.2 | 74.1 ± 1.7 | 81.9 ± 2.3 | 60.0 ± 2.2 |
| 89.1 ± 1.4 | 76.1 ± 1.5 | 96.0 ± 1.1 | 85.8 ± 0.8 | 76.3 ± 1.1 | 62.5 ± 1.4 | 95.3 ± 0.7 | 84.5 ± 1.1 | 87.4 ± 1.2 | 68.3 ± 1.9 | 86.8 ± 1.5 | 74.8 ± 2.2 | 76.8 ± 2.2 | 65.8 ± 3.4 | |
| 92.7 ± 1.1 | 82.5 ± 2.5 | 96.5 ± 0.6 | 78.5 ± 1.2 | 78.5 ± 2.5 | 67.1 ± 1.1 | 96.9 ± 0.8 | 94.3 ± 0.8 | 88.1 ± 1.6 | 73.3 ± 3.0 | 87.6 ± 1.9 | 74.8 ± 2.7 | 78.2 ± 2.8 | 53.3 ± 2.4 | |
| Gaussian SVM | 79.1 ± 1.6 | 69.1 ± 1.2 | 86.9 ± 1.9 | 82.8 ± 2.6 | 54.0 ± 2.9 | 56.7 ± 2.7 | 90.1 ± 3.5 | 87.2 ± 2.2 | 76.1 ± 1.6 | 77.4 ± 2.8 | 73.0 ± 1.4 | 73.1 ± 2.6 | 53.4 ± 1.6 | 71.0 ± 1.4 |
| 78.6 ± 1.3 | 77.6 ± 0.8 | 84.3 ± 1.5 | 83.3 ± 0.1 | 46.3 ± 1.7 | 49.1 ± 1.1 | 81.2 ± 1.2 | 82.8 ± 1.1 | 66.4 ± 1.8 | 53.5 ± 0.9 | 68.7 ± 1.8 | 68.3 ± 0.1 | 49.1 ± 1.9 | 45.8 ± 1.4 | |
| 76.9 ± 0.9 | 83.1 ± 0.5 | 91.6 ± 1.3 | 85.0 ± 0.1 | 48.4 ± 2.2 | 50.3 ± 1.0 | 90.5 ± 1.6 | 93.3 ± 0.1 | 71.2 ± 0.8 | 52.3 ± 1.1 | 73.1 ± 1.5 | 63.3 ± 1.1 | 54.1 ± 1.9 | 47.6 ± 1.1 | |
| kNN | 95.6 ± 1.2 | 75.0 ± 1.2 | 98.1 ± 0.5 | 88.0 ± 2.1 | 86.1 ± 2.1 | 51.1 ± 3.6 | 97.2 ± 0.9 | 85.6 ± 1.3 | 93.7 ± 1.1 | 72.0 ± 1.9 | 93.6 ± 0.4 | 67.9 ± 3.1 | 89.4 ± 1.2 | 59.1 ± 2.7 |
| 93.5 ± 0.8 | 82.3 ± 1.1 | 97.6 ± 0.5 | 88.3 ± 0.7 | 83.9 ± 2.4 | 58.1 ± 3.0 | 96.7 ± 0.9 | 83.5 ± 0.5 | 92.2 ± 1.4 | 75.0 ± 2.3 | 93.8 ± 1.5 | 72.5 ± 1.1 | 85.0 ± 1.3 | 49.6 ± 2.3 | |
| 94.1 ± 0.5 | 87.8 ± 1.3 | 98.1 ± 0.5 | 88.1 ± 0.9 | 85.8 ± 1.8 | 62.8 ± 1.7 | 97.2 ± 0.4 | 94.6 ± 0.5 | 91.4 ± 0.8 | 65.5 ± 1.7 | 93.4 ± 1.4 | 65.5 ± 1.7 | 86.6 ± 1.3 | 60.5 ± 2.6 | |
VA: validation accuracy, TA: training accuracy
White (top row in each cell): relaxation; light gray (middle cellular row): visual task; dark gray (bottom row): mental recall
Testing accuracies ≥ 90% are emboldened.