Table 7.
Classification of the confusion matrix on the UCI-HAR using the proposed model, CNN-BiLSTM, as a feature extractor and SVM as a classifier.
| Activities | Walk | Up | Down | Sit | Std | Lay | Precision | F1 |
|---|---|---|---|---|---|---|---|---|
| Walk | 494 | 2 | 0 | 0 | 0 | 0 | 0.99 | 0.99 |
| Up | 0 | 465 | 5 | 1 | 0 | 0 | 0.98 | 0.99 |
| Down | 0 | 4 | 415 | 0 | 1 | 0 | 0.98 | 0.99 |
| Sit | 0 | 0 | 0 | 448 | 43 | 0 | 0.91 | 0.93 |
| Std | 0 | 0 | 0 | 24 | 508 | 0 | 0.95 | 0.94 |
| Lay | 0 | 0 | 0 | 0 | 0 | 537 | 1.00 | 1.00 |
| Recall | 1.00 | 0.98 | 0.98 | 0.94 | 0.92 | 1.00 | ||
| Accuracy | 97.28% | |||||||
| Kappa | 0.96 | |||||||
The result marked in bold refers to the results that are achieve the best classification of activities using different metrics.