Table 1 Prediction accuracies of LDA, modified KNN and LGBM classification algorithms at baseline 30–32 s for all subtracted SNNr cubes
Prediction 30–32 s | ||||||
|---|---|---|---|---|---|---|
I–O connection | 32 Channels trained | 5 Channels trained | ||||
LDA | KNN | LGBM | LDA | KNN | LGBM | |
1471 reservoir + I-O | 59.4% | 65.60% | 68.8% | 53.1% | 67.20% | 53.1% |
32 | 62.5% | 67.20% | 62.5% | N/A | N/A | N/A |
5 | 54.7% | 60.90% | 54.7% | 59.4% | 73.40% | 59.4% |
P4 | 48.4% | 65.60% | 57.8% | 46.9% | 51.60% | 51.6% |
Fz | 50.0% | 59.40% | 46.9% | 57.8% | 70.30% | 70.3% |
Cz | 43.8% | 60.90% | 57.8% | 39.1% | 60.90% | 54.7% |
PO3 | 0.00% | 57.80% | 48.4% | 0.00% | 57.80% | 54.7% |
F3 | 53.1% | 62.50% | 62.5% | 53.1% | 64.10% | 64.1% |
Best combo out of 5 | Cz + F3 56.3% | P4, Fz, Cz 75.00% | P4, PO3 64% | Cz + F3 62.5% | Fz, Cz 76.6% | Cz 70.3% |