TABLE 2.

Models Performance under Level-1, -2, and -3.

	Data	Models	Youden	Accuracy	Recall	Specificity	F-score	AUC
Level-1	Raw	Logistic regression	22.4%	61.0%	62.8%	59.6%	0.581%	0.699%
		Random forest	23.6%	61.0%	67.4%	56.1%	0.598%	0.675%
		SVM	24.3%	64.0%	48.8%	75.4%	0.538%	0.682%
		XGBoost	28.3%	64.0%	65.1%	63.2%	0.609%	0.686%
Level-2	Imputed	Logistic regression	41.7%	71.0%	69.8%	71.9%	0.674%	0.789%
		Random forest	50.3%	73.0%	90.7%	59.6%	0.742%	0.816%
		SVM	59.1%	78.0%	90.7%	68.4%	0.780%	0.839%
	Raw	XGBoost	57.9%	77.0%	93.0%	64.9%	0.777%	0.809%
Level-3	Imputed	Logistic regression	53.2%	75.0%	88.4%	64.9%	0.752%	0.843%
		Random forest	55.6%	76.0%	90.7%	64.9%	0.765%	0.870%
		SVM	53.9%	76.0%	83.7%	70.2%	0.750%	0.827%
	Raw	XGBoost	63.8%	80.0%	95.3%	68.4%	0.804%	0.847%

Boldfaced Youden index means the best results achieved within specific level of modeling.