Table 3 Model performance for predicting the risk of severely injured (NISS > 15)
Metric | Dataset | LR | RF | XGBoost | SVM | ANN |
|---|---|---|---|---|---|---|
Accuracy [%] | A | 86.8 ± 0.33 | 85.8 ± 0.47 | 87.0 ± 0.38 | 87.1 ± 0.33 | 86.5 ± 0.52 |
B | 86.1 ± 0.40 | 85.1 ± 0.64 | 86.5 ± 0.35 | 86.5 ± 0.47 | 85.5 ± 0.55 | |
C | 86.3 ± 0.36 | 85.1 ± 0.37 | 86.6 ± 0.42 | 86.7 ± 0.32 | 86.0 ± 0.23 | |
D | 86.1–86.2 ± 0.39 | 85.0–85.2 ± 0.70 | 86.5–86.6 ± 0.46 | 86.5–86.6 ± 0.48 | 85.7–86.0 ± 0.42 | |
F1–score [%] | A | 40.0 | 41.9 | 43.2 | 35.9 | 44.9 |
B | 47.1 | 47.9 | 51.2 | 46.1 | 49.8 | |
C | 43.0 | 43.8 | 46.2 | 39.9 | 46.4 | |
D | 47.4–47.7 | 48.2–48.6 | 51.0–51.4 | 46.8–47.1 | 50.7–51.2 | |
AUC | A | 0.86 ± 0.01 | 0.82 ± 0.01 | 0.87 ± 0.01 | 0.83 ± 0.01 | 0.85 ± 0.01 |
B | 0.87 ± 0.01 | 0.84 ± 0.01 | 0.88 ± 0.01 | 0.85 ± 0.01 | 0.86 ± 0.01 | |
C | 0.86 ± 0.01 | 0.82 ± 0.01 | 0.87 ± 0.01 | 0.83 ± 0.01 | 0.85 ± 0.01 | |
D | 0.870–0.870 ± 0.01 | 0.840–0.841 ± 0.01 | 0.880–0.882 ± 0.01 | 0.847–0.850 ± 0.01 | 0.870–0.871 ± 0.01 | |
AUCPR | A | 0.52 ± 0.02 | 0.44 ± 0.02 | 0.53 ± 0.03 | 0.52 ± 0.02 | 0.51 ± 0.02 |
B | 0.59 ± 0.02 | 0.52 ± 0.03 | 0.62 ± 0.02 | 0.60 ± 0.02 | 0.58 ± 0.02 | |
C | 0.55 ± 0.02 | 0.47 ± 0.01 | 0.57 ± 0.02 | 0.55 ± 0.02 | 0.53 ± 0.02 | |
D | 0.595–0.597 ± 0.02 | 0.517–0.522 ± 0.03 | 0.615–0.620 ± 0.02 | 0.59–0.60 ± 0.02 | 0.589–0.591 ± 0.02 |