Table 1. Selection frequency and test accuracy, precision and recall of best model out of the three classifiers and best model of each classifier, obtained with and without FSS application over the individual features of each activity.
| Selected |
Selected k-NN |
Selected LR |
Selected SVM |
|||
|---|---|---|---|---|---|---|
|
Endurance task |
Without FSS | Accuracy | 0.563 | 0.646 | 0.667 | 0.542 |
| Precision | 0.579 | 0.706 | 0.667 | 0.533 | ||
| Recall | 0.458 | 0.500 | 0.667 | 0.667 | ||
| Sel. Frequency | – | 6/48 | 42/48 | 0/48 | ||
| With BB | Accuracy | 0.688 | 0.458 | 0.688 | 0.646 | |
| Precision | 0.645 | 0.458 | 0.655 | 0.629 | ||
| Recall | 0.833 | 0.458 | 0.792 | 0.708 | ||
| Sel. Frequency | – | 5/48 | 42/48 | 1/48 | ||
| With RFE | Accuracy | 0.542 | 0.563 | 0.583 | 0.583 | |
| Precision | 0.550 | 0.579 | 0.571 | 0.571 | ||
| Recall | 0.458 | 0.458 | 0.667 | 0.667 | ||
| Sel. Frequency | – | 36/48 | 6/48 | 6/48 | ||
|
Valsava Maneuver |
Without FSS | Accuracy | 0.583 | 0.417 | 0.583 | 0.542 |
| Precision | 0.583 | 0.389 | 0.583 | 0.545 | ||
| Recall | 0.583 | 0.292 | 0.583 | 0.500 | ||
| Sel. Frequency | – | 0/48 | 43/48 | 5/48 | ||
| With BB | Accuracy | 0.500 | 0.438 | 0.583 | 0.521 | |
| Precision | 0.500 | 0.421 | 0.611 | 0.533 | ||
| Recall | 0.333 | 0.333 | 0.458 | 0.333 | ||
| Sel. Frequency | – | 8/48 | 28/48 | 12/48 | ||
| With RFE | Accuracy | 0.563 | 0.563 | 0.708 | 0.604 | |
| Precision | 0.579 | 0.579 | 0.778 | 0.647 | ||
| Recall | 0.458 | 0.458 | 0.583 | 0.458 | ||
| Sel. Frequency | – | 40/48 | 5/48 | 3/48 | ||
|
Maximum Contraction |
Without FSS | Accuracy | 0.604 | 0.542 | 0.604 | 0.604 |
| Precision | 0.593 | 0.542 | 0.593 | 0.600 | ||
| Recall | 0.667 | 0.542 | 0.667 | 0.625 | ||
| Sel. Frequency | – | 0/48 | 48/48 | 0/48 | ||
| With BB | Accuracy | 0.500 | 0.500 | 0.562 | 0.562 | |
| Precision | 0.500 | 0.500 | 0.560 | 0.556 | ||
| Recall | 0.500 | 0.500 | 0.583 | 0.625 | ||
| Sel. Frequency | – | 40/48 | 6/48 | 2/48 | ||
| With RFE | Accuracy | 0.563 | 0.521 | 0.604 | 0.563 | |
| Precision | 0.565 | 0.522 | 0.593 | 0.556 | ||
| Recall | 0.542 | 0.500 | 0.667 | 0.625 | ||
| Sel. Frequency | – | 38/48 | 10/48 | 0 | ||
|
Wave Task |
Without FSS | Accuracy | 0.625 | 0.521 | 0.667 | 0.562 |
| Precision | 0.625 | 0.533 | 0.667 | 0.565 | ||
| Recall | 0.625 | 0.333 | 0.667 | 0.542 | ||
| Sel. Frequency | – | 0/48 | 41/48 | 7/48 | ||
| With BB | Accuracy | 0.708 | 0.667 | 0.688 | 0.729 | |
| Precision | 0.708 | 0.700 | 0.680 | 0.762 | ||
| Recall | 0.708 | 0.583 | 0.708 | 0.667 | ||
| Sel. Frequency | – | 6/48 | 37/48 | 5/48 | ||
| With RFE | Accuracy | 0.771 | 0.792 | 0.729 | 0.667 | |
| Precision | 0.842 | 0.889 | 0.704 | 0.654 | ||
| Recall | 0.667 | 0.667 | 0.792 | 0.704 | ||
| Sel. Frequency | – | 43/48 | 4/48 | 1/48 |
Notes.
- k-NN
- k-nearest neighbors
- LR
- logistic regression
- SVM
- support vector machine
- FSS
- feature subset search
- BB
- branch and bound
- RFE
- recursive feature elimination