Table 3 Comparison of the three ML models by metric.

From: Predicting low density lipoprotein cholesterol target attainment using machine learning in patients with coronary artery disease receiving moderate-dose statin therapy

Dataset

Metrics

Value [95% CI]

  

XGBoost

RF

LR

Training

(n = 6,581)

ACC

0.642

[0.638–0.645]

0.589

[0.584–0.597]

0.584

[0.578–0.590]

SEN

0.639

[0.627–0.651]

0.897

[0.886–0.912]

0.892

[0.882–0.902]

SPE

0.644

[0.628–0.661]

0.304

[0.295–0.311]

0.298

[0.291–0.307]

PPV

0.625

[0.619–0.632]

0.544

[0.541–0.549]

0.541

[0.538–0.545]

NPV

0.658

[0.655–0.661]

0.762

[0.745–0.789]

0.748

[0.730–0.766]

AUROC

0.702

[0.694–0.712]

0.690

[0.679–0.704]

0.686

[0.676–0.695]

Test

(n = 2,821)

ACC

0.659

[0.641–0.677]

0.576

[0.557–0.595]

0.582

[0.565–0.602]

SEN

0.661

[0.636–0.685]

0.887

[0.868–0.905]

0.894

[0.878–0.910]

SPE

0.658

[0.632–0.684]

0.286

[0.263–0.310]

0.293

[0.268–0.316]

PPV

0.642

[0.617–0.669]

0.536

[0.516–0.557]

0.540

[0.520–0.563]

NPV

0.676

[0.650–0.700]

0.733

[0.692–0.770]

0.749

[0.713–0.787]

AUROC

0.706

[0.686–0.726]

0.688

[0.667–0.708]

0.681

[0.661–0.701]

  1. Abbreviations: ACC, accuracy; AUROC, area under the receiver operating characteristic; LR, logistic regression; ML, machine learning; NPV, negative predictive value; PPV, positive predictive value; RF, random forest; SEN, sensitivity; SPE, specificity; XGBoost, extreme gradient boosting.
Back to article page