Table 3.

ML applications in perinatal death research.

Reference	ML application	Input	ML technique	Main output
(62)	Complication prediction	Clinical parameters	Extreme GB	AUC = 0.842, Ac = 94.71%, Se = 45.3%, Sp = 95%, PPV = 4.81%, NPV = 99.68%, +LR = 9.03 and -LR = 0.58 for stillbirth
(63)	Complication prediction	Clinical parameters	LR	AUC = 0.82 for stillbirth
(64)	Complication prediction	Clinical parameters	SNNN	AUC = 0.76, Se = 38% and Sp = 90% for early stillbirth
(65)	Complication prediction	Clinical parameters	LR	AUC = 0.872 for neonatal death
(66)	Complication prediction	Clinical parameters	RF	AUC = 0.79, Ac = 0.87, Se = 0.54, Sp = 0.88, PPV = 0.15 and NPV = 0.98
(67)	Complication prediction	Clinical parameters	DT, GB, LR, RF and SVM	AUC = 90.00%, Ac = 90.56%, Se = 91.37%, Sp = 88.10%, Pr = 88.02% and F1 = 90.58% for stillbirth before delivery and during labor
(68)	Complication prediction	Clinical parameters	MLP	AUC = 95.99%, Ac = 96.79%, Se = 86.20%, Sp = 98.37%, RMSE = 0.1702 and RRSE = 47.47% for neonatal death
(69)	Complication prediction	Clinical parameters	SL	AUC = 0.89 and U = -0.0003 for neonatal death
(70)	Complication prediction	Clinical parameters	RF	AUC = 0.922, Ac = 0.903, Se = 0.674, Sp = 0.919, PPV = 0.377, F1 = 0.477 and mean F1 = 0.712 for neonatal death
(71)	Complication prediction	Clinical parameters	ANN	AUC = 0.92, Ac = 0.86, Se = 0.86, Sp = 0.83, Pr = 0.96 and F1 = 0.91 for neonatal death

ML, machine learning; GB, gradient boosting; LR, logistic regression; SNNN, self-normalizing neural networks; RF, random forest; DT, decision tree; SVM, support vector machines; MLP, multilayer perceptron; SL, SuperLearner; ANN, artificial neural networks; AUC, area under the receiver operating characteristic curve; Ac, accuracy; Se, sensitivity; Sp, specificity; PPV, positive predictive value; NPV, negative predictive value; +LR, positive likelihood ratio; -LR, negative likelihood ratio; Pr, precision; RMSE, root mean squared error; RRSE, root relative squared error.