TABLE 4.
Model performance for gait classification across sensor configurations (setup).
| Setup | Model | Sens | Spec | Acc | Accbal | PPV |
|---|---|---|---|---|---|---|
| All | SVM | 90.7 ± 7.5 | 94.5 ± 4.2 | 93.3 ± 4.3 | 92.6 ± 5.5 | 85.6 ± 15.3 |
| LR | 92.1 ± 6.1 | 93.7 ± 4.8 | 93.1 ± 4.6 | 92.9 ± 5.1 | 84.5 ± 16 | |
| kNN | 89.5 ± 6.3 | 94.4 ± 4.6 | 92.7 ± 4.3 | 91.9 ± 4.9 | 85.5 ± 15.4 | |
| No chest | SVM | 89.8 ± 7.3 | 94.2 ± 4.6 | 92.7 ± 4.5 | 92.0 ± 5.5 | 85.0 ± 15.7 |
| LR | 91.4 ± 6.5 | 93.5 ± 4.9 | 92.7 ± 4.7 | 92.5 ± 5.3 | 84.1 ± 15.9 | |
| kNN | 88.5 ± 6.9 | 94.2 ± 4.5 | 92.3 ± 4.3 | 91.3 ± 5.3 | 85.0 ± 15.5 | |
| Non affected | SVM | 89.2 ± 7.2 | 94.2 ± 4 | 92.5 ± 3.9 | 91.7 ± 5 | 84.8 ± 15.1 |
| LR | 90.6 ± 6.2 | 93.1 ± 5.1 | 92.0 ± 4.8 | 91.8 ± 4.9 | 83.0 ± 16.8 | |
| kNN | 86.9 ± 6.6 | 94.0 ± 4.2 | 91.5 ± 4 | 90.4 ± 4.9 | 84.4 ± 15 | |
| Affected | SVM | 89.5 ± 7.1 | 92.6 ± 4 | 91.6 ± 4.1 | 91.0 ± 5.1 | 82.0 ± 15.2 |
| LR | 90.4 ± 6.8 | 92.1 ± 4.9 | 91.3 ± 4.6 | 91.3 ± 5.2 | 81.3 ± 16.4 | |
| kNN | 86.4 ± 8.1 | 93.7 ± 4.5 | 91.2 ± 4.3 | 90.0 ± 5.4 | 83.9 ± 15.5 | |
| Wrists-only | SVM | 79.9 ± 9.4 | 90.6 ± 4.3 | 87.1 ± 4.5 | 85.3 ± 6.1 | 76.7 ± 16.9 |
| LR | 82.8 ± 8 | 86.6 ± 5.4 | 84.9 ± 5.1 | 84.7 ± 5.5 | 71.8 ± 17.1 | |
| kNN | 80.0 ± 9.1 | 83.8 ± 4.6 | 82.4 ± 5.1 | 81.9 ± 5.9 | 67.5 ± 16.4 |
Mean and standard deviations of performance measures in %. Accbal, balanced accuracy; kNN, k-Nearest Neighbor; LR, logistic regression; PPV, positive predictive value.