TABLE 3.

Classification performance of the SVM model (radial basis kernel) and RF model.

		SVM model (radial basis kernel)			RF model
Group	Method	ACC	SEN	SPE	ACC	SEN	SPE
Validation dataset	Fisher score	90.23% ± 4.78%	85.91 ± 8.94%	93.71 ± 5.45%	87.07 ± 5.53%	80.29 ± 10.53%	92.46 ± 5.85%
Test dataset	Fisher score	84.48 ± 5.58%	82.87 ± 12.56%	85.98 ± 7.11%	83.19 ± 5.89%	77.80 ± 12.38%	88.19 ± 6.47%
Validation dataset	Lasso	95.90 ± 3.29%	92.82 ± 6.84%	98.26 ± 2.80%	90.81 ± 4.76%	84.58 ± 9.13%	95.74 ± 4.60%
Test dataset	Lasso	88.94 ± 5.33%	80.56 ± 10.71%	96.70 ± 3.83%	83.68 ± 6.85%	74.26 ± 13.86%	92.41 ± 5.78%
Validation dataset	mRMR	93.00 ± 4.19%	89.07 ± 7.11%	96.08 ± 3.98%	90.28 ± 4.97%	84.14 ± 10.03%	95.08 ± 4.96%
Test dataset	mRMR	86.08 ± 5.68%	79.76 ± 12.13%	91.93 ± 5.00%	83.52 ± 6.62%	75.85 ± 13.47%	90.63 ± 6.01%

Under the sample disturbance of five-fold cross-validation, we carried out three different kinds of composite function disturbances separately to screen features in the training dataset and repeated the process 100 times. The retained features were incorporated into the SVM model and RF model each time, and we then calculated the models’ classification performance in the validation dataset and test dataset separately. The measures are presented as mean ± standard deviation. SVM, support vector machine; RF, random forest; ACC, accuracy; SEN, sensitivity; SPE, specificity; Lasso, least absolute shrinkage and selection operator; mRMR, max-relevance and min-redundancy.