Development and external validation of a machine learning model for brain injury in pediatric patients on extracorporeal membrane oxygenation

Table 3 Performance of internal validation models (using data balanced by ADASYN method)

Model	AUC	AUPRC	Accuracy	TPR	TNR	FPR	FNR	PPV	NPV	F1	AIC	Brier Score	PLR	NLR
RF	0.912	0.670	0.879	0.900	0.700	0.037	0.557	0.963	0.443	0.931	190.753	0.108	14.551	0.555
XGB	0.867	0.423	0.791	0.800	0.720	0.038	0.710	0.962	0.290	0.873	244.613	0.137	5.622	0.756
LGB	0.840	0.433	0.804	0.825	0.620	0.050	0.713	0.950	0.287	0.883	257.834	0.135	7.675	0.722
DT	0.655	0.184	0.712	0.736	0.500	0.072	0.823	0.928	0.177	0.821	312.022	0.202	2.350	0.901
SVM	0.810	0.372	0.777	0.795	0.620	0.052	0.744	0.948	0.256	0.865	266.943	0.160	5.177	0.783
NaiveBayes	0.704	0.257	0.816	0.861	0.420	0.071	0.744	0.929	0.256	0.894	297.572	0.147	4.408	0.747
NN	0.835	0.334	0.779	0.790	0.680	0.044	0.730	0.956	0.270	0.865	258.970	0.165	6.444	0.763
GLM	0.718	0.288	0.697	0.708	0.600	0.060	0.810	0.940	0.190	0.808	294.173	0.200	3.113	0.863
GBM	0.686	0.203	0.808	0.856	0.380	0.076	0.768	0.924	0.232	0.889	308.991	0.243	2.547	0.910
AdaBoost	0.789	0.360	0.785	0.806	0.600	0.053	0.739	0.947	0.261	0.871	279.334	0.151	4.342	0.799

AUC Area under the curve; AUPRC: the area under the precision-recall curve; TPR: True positive rate; TNR: True negative rate; FPR: False positive rate; FNR: False negative rate; PPV: Positive predictive value; NPV: Negative predictive value; AIC: Akaike Information Criterion; PLR: Positive likelihood ratio; NLR: Negative likelihood ratio

ISSN: 1364-8535