Table 7.
Performance (mean and SD) of task-specific quality classification models when evaluated on 3-class ordinal classification.
| Task | Best model | QWKa | Accuracy | OCAb |
| Finger-tapping | Random forest | 0.71 (0.06) | 70.0 (4.1) | 82.1 (2.7) |
| Smile | LightGBM | 0.56 (0.07) | 58.9 (4.9) | 76.9 (2.9) |
| Speech | Random forest | 0.72 (0.09) | 81.8 (3.9) | 89.9 (2.2) |
aQWK: quadratic weighted Cohen κ.
bOCA: ordinal classification accuracy.