Table 6 Top 10 features selected by each method and the optimum number selected by Kneedle algorithm (corresponding accuracy)
From: Feature ranking and rank aggregation for automatic sleep stage classification: a comparative study
| Â | ReliefF | Fisher | Chi2 | IG | CMIM | MRMR-MID | MRMR-MIQ | Borda | RRA |
|---|---|---|---|---|---|---|---|---|---|
Top 10 features | F28 | F36 | F35 | F9 | F15 | F35 | F35 | F36 | F36 |
F36 | F35 | F9 | F35 | F36 | F39 | F42 | F35 | F35 | |
F7 | F31 | F11 | F11 | F9 | F36 | F15 | F9 | F9 | |
F49 | F9 | F31 | F31 | F8 | F22 | F36 | F31 | F31 | |
F41 | F29 | F36 | F36 | F1 | F15 | F22 | F22 | F27 | |
F27 | F11 | F27 | F4 | F34 | F31 | F23 | F27 | F22 | |
F20 | F25 | F26 | F27 | F35 | F29 | F31 | F29 | F17 | |
F23 | F27 | F4 | F26 | F28 | F23 | F38 | F11 | F29 | |
F6 | F12 | F25 | F25 | F6 | F9 | F29 | F15 | F11 | |
F22 | F22 | F14 | F29 | F48 | F38 | F9 | F20 | F20 | |
MLFN | 7 (0.75) | 5 (0.76) | 7 (0.76) | 7 (0.76) | 3 (0.74) | 5 (0.76) | 5 (0.76) | 5 (0.76) | 7 (0.77) |
k-NN | 7 (0.69) | 5 (0.71) | 9 (0.73) | 9 (0.73) | 3 (0.68) | 7 (0.75) | 11 (0.75) | 9 (0.74) | 7 (0.73) |