Table 2.

Model performance of four classifiers on the validation set; a random forest wrapper method (Boruta), regression partition trees (Rpart), LASSO, and extreme gradient boosting (XGBoost).

	Random forest	Rpart	LASSO	XGBoost	Ensemble
miRNAs selected by model, n	10	4	13	8	20
Sensitivity (95% CI)	0.86 (0.65-0.97)	0.91 (0.71-0.99)	0.77 (0.55-0.92)	0.91 (0.71-0.99)	0.91 (0.71-0.99)
Specificity (95% CI)	0.71(0.42-0.92)	0.64 (0.35-0.87)	0.64 (0.35-0.87)	0.71 (0.42-0.92)	0.64 (0.35-0.87)
Positive predictive value (95% CI)	0.83 (0.61-0.95)	0.80 (0.59-0.93)	0.77 (0.55-0.92)	0.83 (0.63-0.95)	0.80 (0.59-0.93)
Negative predictive value (95% CI)	0.77 (0.46-0.95)	0.82 (0.48-0.92)	0.64 (0.35-0.86)	0.83 (0.52-0.98)	0.82 (0.48-0.92)
Correct classification rate (95% CI)	0.81 (0.64-0.92)	0.81 (0.64-0.92)	0.72 (0.55-0.86)	0.83 (0.67-0.94)	0.81 (0.64-0.92)
AUC (95% CI)	0.84 (0.69-1)	0.79 (0.63-0.95)	0.79 (0.63-0.94)	0.82 (0.66-0.99)	0.85 (0.70-1)