Table 3. ML results for each classifier. Soft voting comprised of a combination of the four ML algorithms.

Classification algorithm	Training set					Internal validation set
	AUC	Accuracy	Recall	Precision		AUC	Accuracy	Recall	Precision
Linear Support Vector Machine	0.906±0.020	0.857±0.021	0.857±0.021	0.857±0.021		0.766±0.059	0.711±0.056	0.711±0.056	0.711±0.056
Random Forest	0.901±0.018	0.823±0.024	0.823±0.024	0.823±0.024		0.922±0.034	0.858±0.045	0.858±0.046	0.858±0.046
Adaptive Boost	0.894±0.020	0.810±0.024	0.811±0.024	0.810±0.024		0.882±0.043	0.816±0.051	0.817±0.051	0.816±0.050
Extreme Gradient Boost	0.949±0.013	0.877±0.021	0.880±0.021	0.878±0.021		0.910±0.037	0.833±0.047	0.833±0.046	0.833±0.046
Soft Voting Classifier	0.921±0.016	0.837±0.023	0.837±0.023	0.837±0.023		0.890±0.040	0.820±0.046	0.822±0.046	0.820±0.046

Data are shown as performance value ± 95% confidence interval. ML, machine learning; AUC, area under the receiver operating characteristic curve.