Table 4. Comparison of the predictive performance of four machine learning models on testing data.

Model		XGBoost	Boosted Tree	Bootstrap Forest	Neural Networks	4 models mean	PLS-DA
Full model	No. of variables	530	530	530	530	-	530
	Accuracy	0.9688	0.9201	0.8889	0.8125	0.8976	0.7326
	Misclassification	0.0313	0.0799	0.1111	0.1875	0.1025	0.2674
	AUC	0.9661	0.9176	0.8801	0.8086	0.8931	0.7297
	Sensitivity	0.9877	0.9383	0.9506	0.8395	0.9290	0.7531
	Specificity	0.9444	0.8968	0.8095	0.7778	0.8571	0.7063
	Precision	0.9581	0.9212	0.8652	0.8293	0.8934	0.7673
Reduced model	No. of variables	28	33	15	195	-	311
	Accuracy	0.9722	0.9653	0.8576	0.8299	0.9063	0.7431
	Misclassification	0.0278	0.0347	0.1424	0.1701	0.0938	0.2569
	AUC	0.9735	0.9647	0.8514	0.8258	0.9039	0.7381
	Sensitivity	0.9630	0.9691	0.9012	0.8580	0.9228	0.7778
	Specificity	0.9841	0.9603	0.8016	0.7937	0.8849	0.6984
	Precision	0.9873	0.9691	0.8538	0.8424	0.9132	0.7683