Table 2. The performance of classification models based on RCSP-set-Threshold (28 genes) developed using different machine learning techniques on training and independent or external validation dataset.

Technique	Dataset	Performance Measures
		Sensitivity	Specificity	Accuracy (%)	MCC	ROC
RF	Training	73.62	72.12	73.03	0.45	0.77
	Validation	73.02	60.98	68.27	0.34	0.74
Naive Bayes	Training	75.98	67.27	72.55	0.43	0.76
	Validation	77.78	60.98	71.15	0.39	0.76
SMO	Training	83.86	55.76	72.79	0.42	0.70
	Validation	80.95	53.66	70.19	0.36	0.67
J48	Training	64.17	66.06	64.92	0.3	0.67
	Validation	68.25	58.54	64.42	0.26	0.67
SVM	Training	75.98	69.09	73.27	0.45	0.78
	Validation	74.6	65.85	71.15	0.4	0.77

These RCSP-set-Threshold features are selected by the threshold-based approach followed by the removal of correlated features.