Table 5.

The average values of 100 times classifying for testing data are shown. Results include the mean value of accuracy, AUC, sensitivity, specificity, precision, and F-score for three classes. Feature selection methods are shown in the first column, including NCA, MRMR, and LASSO. The best values of each evaluation parameter are bolded

	Classifiers	Accuracy	AUC	Sensitivity	Specificity	Precision	F-score
NCA	Cosine KNN	0.82	0.87	0.72	0.85	0.74	0.72
	Fine KNN	0.84	0.83	0.75	0.87	0.78	0.76
	Subspace KNN	0.80	0.84	0.69	0.84	0.74	0.69
	Cross-entropy decision tree	0.76	0.75	0.61	0.82	0.61	0.60
	RUSBoosted trees	0.78	0.76	0.66	0.84	0.66	0.65
	Cubic SVM	0.75	0.80	0.62	0.81	0.65	0.60
	Random forest	0.79	0.88	0.64	0.83	0.67	0.65
MRMR	Cosine KNN	0.60	0.59	0.38	0.69	0.35	0.37
	Fine KNN	0.61	0.57	0.40	0.70	0.39	0.39
	Subspace KNN	0.61	0.56	0.39	0.68	0.37	0.37
	Cross-entropy decision tree	0.64	0.68	0.44	0.72	0.45	0.44
	RUSBoosted trees	0.65	0.62	0.48	0.74	0.47	0.46
	Cubic SVM	0.63	0.57	0.45	0.72	0.42	0.42
	Random forest	0.68	0.71	0.47	0.74	0.48	0.47
LASSO	Cosine KNN	0.72	0.81	0.54	0.78	0.55	0.53
	Fine KNN	0.75	0.69	0.58	0.80	0.60	0.59
	Subspace KNN	0.75	0.74	0.57	0.79	0.61	0.58
	Cross-entropy decision tree	0.75	0.76	0.63	0.81	0.63	0.62
	RUSBoosted trees	0.77	0.80	0.65	0.83	0.65	0.64
	Cubic SVM	0.73	0.74	0.56	0.79	0.57	0.55
	Random forest	0.76	0.81	0.59	0.81	0.60	0.58