Table 3 The performance of machine learning models on test sets.

From: Predicting mortality and risk factors of sepsis related ARDS using machine learning models

Cohort and Models

AUROC

Cutoff

Sensitivity

Specificity

PPV

NPV

XGBoost

0.833(0.804–0.861)

0.377

0.66(0.603–0.717)

0.867(0.843–0.891)

0.636(0.58–0.693)

0.879(0.855–0.902)

LightGBM

0.827(0.798–0.856)

0.139

0.717(0.663–0.771)

0.784(0.755–0.814)

0.54(0.488–0.592)

0.887(0.863–0.911)

RF

0.846(0.818–0.874)

0.319

0.789(0.74–0.838)

0.783(0.753–0.812)

0.562(0.511–0.612)

0.913(0.891–0.935)

NB

0.799(0.768–0.831)

0.973

0.74(0.687–0.792)

0.746(0.715–0.777)

0.506(0.457–0.556)

0.89(0.866–0.915)

CART

0.753(0.718–0.787)

0.155

0.751(0.699–0.803)

0.679(0.646–0.712)

0.452(0.406–0.499)

0.885(0.859–0.911)

LR

0.826(0.796–0.856)

0.257

0.774(0.723–0.824)

0.751(0.72–0.783)

0.523(0.474–0.572)

0.904(0.881–0.927)

  1. AUROC area under the receiver operating characteristic curve, PPV positive predictive value, NPV negative predictive value. NB naive bayes,CART classification and regression tree, LightGBM light gradient boosting machine, LR logistic regression, RF random forest, XGBoost extreme gradient boosting.
Back to article page