Table 4 Performance metrics for 10 candidate model pipelines (Model Numbers from Table 3) classifying daily activity of waterfowl into 8 classes using GPS-derived feature datasets reflecting movement and timing, habitat, and history of movement
Evaluation metric | Model number (%) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | |
Accuracy | 95.2 | 86.4 | 86.7 | 94.8 | 85.8 | 81.8 | 85.5 | 94.8 | 94.9 | 92.4 |
Macro-precision | 96.3 | 76.3 | 80.7 | 96.3 | 76.7 | 70.3 | 75.4 | 96.3 | 96.3 | 86.5 |
Macro-recall | 87.1 | 71.6 | 72.5 | 86.7 | 69.9 | 63.4 | 70.0 | 86.7 | 87.2 | 82.9 |
Macro-F1 | 89.9 | 73.3 | 74.6 | 89.7 | 72.1 | 65.8 | 71.8 | 89.7 | 89.9 | 84.1 |
Weighted-precision | 95.3 | 85.7 | 86.4 | 94.8 | 85.3 | 80.9 | 84.7 | 94.8 | 94.9 | 92.3 |
Weighted-recall | 95.2 | 86.4 | 86.7 | 94.8 | 85.8 | 81.8 | 85.5 | 94.8 | 94.9 | 92.4 |
Weighted-F1 | 95.0 | 86.0 | 86.3 | 94.6 | 85.4 | 81.1 | 84.9 | 94.6 | 94.7 | 92.2 |