Table 1.
Percentage of machine-learned behaviours automatically classified from wrist-worn accelerometer data.
| Prediction → Ground truth ↓ |
Sleep | Sit/stand | Vehicle | Walking | Mixed-activity | Bicycling |
|---|---|---|---|---|---|---|
| Sleep | 97% | 3% | <1% | <1% | 1% | <1% |
| Sit/stand | 3% | 89% | 1% | 3% | 3% | <1% |
| Vehicle | <1% | 13% | 74% | 3% | 9% | <1% |
| Walking | 1% | 11% | 2% | 71% | 15% | 1% |
| Mixed-activity | 1% | 20% | 2% | 19% | 57% | 1% |
| Bicycling | 1% | 1% | 1% | 12% | 14% | 71% |
Confusion matrix after leave-one-out validation on 84,616 labelled minutes of human activity in free-living environments: the CAPTURE-24 study 2014–2015 (n = 132).