Table 5 Results comparing the labelled contextual features with the derived ML ones
From: Assessing the added value of context during stress detection from wearable data
Model | Accuracy (std) | Cohen Kappa (std) | Weighted F1 (std) | Compared to dummy (std) | |
|---|---|---|---|---|---|
Physio (baseline) | 40.24 (1.60) | 2.05 (1.10) | 40.58 (0.81) | 5.26 (2.44) | |
Physio + activity | Labelled | 38.79 (1.47) | − 0.94 (0.51) | 39.22 (0.75) | 3.90 (4.00) |
| Â | Predicted | 43.40 (2.16) | 1.14 (1.02) | 42.19 (1.54) | 6.86 (1.70) |
Physio + sleep | Labelled | 43.96 (1.59) | 4.03 (1.25) | 43.18 (0.41) | 7.85 (2.83) |
| Â | Predicted | 44.20 (2.97) | 3.00 (2.89) | 43.06 (2.42) | 7.74 (0.83) |
Physio + activity + sleep | Labelled | 42.09 (5.08) | 3.42 (4.02) | 41.88 (4.17) | 6.55 (0.92) |
| Â | Predicted | 41.69 (4.35) | 1.49 (0.78) | 41.48 (2.93) | 6.15 (0.31) |