Table 3.

EGFR mutation prediction results on Test dataset, base classifiers.

Feature Selection	Classifier	SMOTE	Accuracy	Sensitivity	Specificity	AUC
MW (5 features)	nnet	No	0.83	0.00	0.98	0.43
ReliefF (15 features)	SVM	Yes	0.76	0.66	0.78	0.68
ReliefF (10 features)	RF	Yes	0.76	0.41	0.82	0.67
ReliefF (15 features)	nnet	Yes	0.76	0.58	0.79	0.67
ReliefF (5 features)	RF	Yes	0.77	0.50	0.82	0.64
ReliefF (20 features)	RF	Yes	0.73	0.16	0.83	0.63
ReliefF (20 features)	SVM	Yes	0.68	0.66	0.69	0.63
ReliefF (5 features)	nnet	Yes	0.71	0.50	0.75	0.60
ReliefF (5 features)	SVM	Yes	0.79	0.25	0.89	0.59
ReliefF (15 features)	RF	Yes	0.72	0.25	0.80	0.57
MW (5 features)	gbm	Yes	0.68	0.16	0.78	0.53