Table 6 Performance metrics at different thresholds for XGBoost model.

From: Development and validation of a machine learning model for critical progression risk in pediatric severe community-acquired pneumonia

阈值

Sensitivity

Specificity

假阳性率(FPR)

阳性预测值(PPV)

阴性预测值(NPV)

F1-score

0.10

0.000

1.000

0.000

-

0.317

-

0.15

0.000

1.000

0.000

-

0.317

-

0.20

0.000

1.000

0.000

-

0.317

-

0.25

0.326

1.000

0.000

1.000

0.408

0.491

0.30

0.698

1.000

0.000

1.000

0.606

0.822

0.35

0.767

1.000

0.000

1.000

0.667

0.868

0.40

0.884

1.000

0.000

1.000

0.800

0.938

0.45

0.884

0.900

0.100

0.950

0.783

0.916

0.50

0.977

0.750

0.250

0.894

0.938

0.933

0.55

1.000

0.550

0.450

0.827

1.000

0.905

0.60

1.000

0.300

0.700

0.754

1.000

0.860

0.65

1.000

0.050

0.950

0.694

1.000

0.819

0.70

1.000

0.000

1.000

0.683

-

0.811

0.75

1.000

0.000

1.000

0.683

-

0.811

0.80

1.000

0.000

1.000

0.683

-

0.811

0.85

1.000

0.000

1.000

0.683

-

0.811

0.90

1.000

0.000

1.000

0.683

-

0.811

  1. The performance metrics of the XGBoost model at various classification thresholds. Sensitivity, specificity, false positive rate, positive predictive value (PPV), negative predictive value (NPV), and F1 score are reported for each threshold. The table highlights the trade-offs between sensitivity and specificity and demonstrates the optimized performance at a threshold of 0.45, where the false positive rate is significantly reduced while maintaining high sensitivity.
Back to article page