Table 8.

Model performance comparison for the extreme low birthweight prediction task.

Method and model		Average accuracy	Average precision	Average recall	Average F1-score	Average AUROCa
Before rebalancing
	LRb	0.93	0.13	0.17	0.15	0.58
	NBc	0.87	0.05	0.45	0.09	0.66
	RFd	0.95	0.17	0.66	0.27	0.81
	XGBooste	0.97	0.14	0.18	0.16	0.58
	AdaBoostf	0.97	0.14	0.18	0.16	0.58
	MLPg	0.97	0.14	0.18	0.16	0.58
	Sequential ANNh	0.98	0.15	0.02	0.04	0.51
Random undersampling
	LR	0.74	0.04	0.80	0.08	0.77
	NB	0.87	0.05	0.46	0.10	0.67
	RF	0.74	0.04	0.77	0.08	0.75
	XGBoost	0.75	0.05	0.80	0.09	0.77
	AdaBoost	0.75	0.04	0.79	0.08	0.77
	MLP	0.71	0.04	0.72	0.07	0.72
	Sequential ANN	0.69	0.04	0.82	0.07	0.76
Random oversampling
	LR	0.75	0.04	0.80	0.08	0.77
	NB	0.88	0.05	0.46	0.10	0.67
	RF	0.96	0.04	0.06	0.05	0.52
	XGBoost	0.76	0.05	0.78	0.09	0.77
	AdaBoost	0.76	0.05	0.78	0.09	0.77
	MLP	0.71	0.05	0.71	0.08	0.72
	Sequential ANN	0.80	0.04	0.57	0.08	0.69
SMOTEi
	LR	0.80	0.05	0.68	0.09	0.74
	NB	0.82	0.05	0.59	0.09	0.71
	RF	0.98	0.11	0.06	0.08	0.53
	XGBoost	0.94	0.07	0.30	0.12	0.62
	AdaBoost	0.85	0.05	0.57	0.10	0.71
	MLP	0.94	0.06	0.21	0.09	0.58
	Sequential ANN	0.88	0.05	0.37	0.08	0.63
Weight rebalancing
	LR	0.71	0.04	0.77	0.07	0.74
	NB	0.71	0.04	0.77	0.07	0.74
	RF	0.89	0.01	0.80	0.18	0.85
	XGBoost	0.61	0.03	0.85	0.06	0.61
	AdaBoost j	0.85	0.07	0.84	0.14	0.84

^aAUROC: area under the receiver operating characteristic curve.

^bLR: logistic regression.

^cNB: naive Bayes.

^dRF: random forest.

^eXGBoost: extreme gradient boosting.

^fAdaBoost: adaptive boosting.

^gMLP: multilayer perceptron.

^hANN: artificial neural network.

ⁱSMOTE: synthetic minority oversampling technique.

^jModel with the best performance is italicized.