Table 4.

Validation and evaluation results of three machine learning classifiers performance

Classifiers	AUC (95% CI)	Se (95% CI)	Sp (95% CI)	PPV	NPV	Correct rate
SVM a	0.996 (0.989, 1.000)	0.982 (0.906, 1.000)	0.907 (0.797, 0.969)	0.918	0.946	0.946
SVM b	0.813 (0.761, 0.866)	0.780 (0.707, 0.842)	0.717 (0.618, 0.803)	0.816	0.756	0.756
RF a	0.995 (0.988, 1.000)	0.983 (0.906, 1.000)	0.907 (0.797, 0.969)	0.919	0.955	0.955
RF b	0.727 (0.665, 0.788)	0.723 (0.647, 0.791)	0.525 (0.422, 0.627)	0.696	0.636	0.636
LR a	0.991 (0.971, 1.000)	0.965 (0.879, 0.996)	0.982 (0.901, 1.000)	0.982	0.973	0.973
LR b	0.783 (0.725, 0.841)	0.516 (0.435, 0.596)	0.869 (0.786, 0.928)	0.859	0.640	0.640

SVM, support vector machine; RF, randomforest; LR, logistic regression; Se, sensitivity; Sp, specificity; PPV, positive predictive value; NPV, negative predictive value; AUC, area under the ROC curve; ROC, receiver operating characteristic curve

^aVerified in the 50% samples of GSE12288 (111/222)

^bVerified in the integrated dataset of GSE7638 and GSE66360 (258)