Table 3. Performance of all models for their corresponding best subsets of observed variables.
| Mode | Model | HSMM | HMM | SVM | RF | ANN |
| Subset |
|
|
|
|
|
|
| Accuracy | 80.3% | 79.1% | 79.0% | 76.4% | 79.2% | |
| F | Recall | 86.6% | 84.7% | 88.3% | 85.5% | 88.5% |
| Precision | 69.3% | 69.4% | 65.8% | 64.5% | 65.7% | |
| F1 | 77.0% | 76.3% | 75.4% | 73.5% | 75.4% | |
| Duration | 4 | 14 | 17 | 29 | 22 | |
| S | Recall | 37.5% | 64.9% | 56.3% | 62.0% | 59.5% |
| Precision | 56.9% | 56.1% | 57.6% | 47.5% | 54.3% | |
| F1 | 66.7% | 60.2% | 56.9% | 53.7% | 56.9% | |
| Duration | 7 | 4 | 33 | 41 | 26 | |
| C | Recall | 91.0% | 87.4% | 89.6% | 76.9% | 88.7% |
| Precision | 87.3% | 86.4% | 71.9% | 72.0% | 75.2% | |
| F1 | 89.1% | 86.9% | 79.8% | 74.3% | 82.1% | |
| Duration | 1 | 2 | 7 | 25 | 7 |
Notes: In bold, the highest values of accuracy and F1. F: fishing; S: searching; C: cruising. Duration values are scaled by (
).