Fig. 6

Sensitivity analysis of XGBoost performance across different classification thresholds. This figure shows the sensitivity analysis of XGBoost performance for various classification thresholds. As the threshold increases, sensitivity increases, specificity decreases, and the false positive rate rises. The analysis identifies an optimal threshold of 0.45, balancing high sensitivity (0.884) and specificity (0.900), while significantly reducing the false positive rate (0.100). This adjustment leads to an improved F1 score (0.916), optimizing classification performance.