Table 5 Class-specific performance of models in switch and trap classification.
From: Machine learning-based analysis of defensive strategies in basketball using player movement data
Model | Class | Accuracy (%) | Precision (%) | Recall (%) | F1-score (%) |
|---|---|---|---|---|---|
Random forest | Switch | 79.2 | 78.0 | 76.4 | 77.2 |
Trap | 81.5 | 80.1 | 79.3 | 79.7 | |
Neither | 85.6 | 83.9 | 82.7 | 83.3 | |
LSTM | Switch | 82.6 | 81.3 | 82.0 | 81.6 |
Trap | 85.4 | 84.2 | 84.8 | 84.5 | |
Neither | 87.3 | 86.0 | 85.5 | 85.7 | |
CNN | Switch | 85.8 | 84.6 | 85.3 | 85.0 |
Trap | 87.9 | 87.2 | 86.5 | 86.8 | |
Neither | 89.2 | 88.0 | 87.7 | 87.9 | |
Hybrid model | Switch | 90.3 | 89.1 | 88.7 | 88.9 |
Trap | 91.7 | 90.9 | 90.6 | 90.7 | |
Neither | 92.8 | 91.5 | 91.2 | 91.4 |
