Table 6. The performance of models developed using different machine techniques based on RCSP-set-Weka-Hall (38 genes) selected from Weka.

Technique	Dataset	Performance Measures
		Sensitivity	Specificity	Accuracy (%)	MCC	ROC
RF	Training	75.10	78.66	76.50	0.53	0.84
	Validation	67.19	78.57	71.70	0.45	0.75
Naive Bayes	Training	79.84	71.34	76.50	0.51	0.83
	Validation	75.00	66.67	71.70	0.41	0.79
SMO	Training	85.77	66.46	78.18	0.54	0.76
	Validation	82.81	59.52	73.58	0.44	0.71
J48	Training	71.54	61.59	67.63	0.33	0.69
	Validation	68.75	71.43	69.81	0.39	0.68
SVM	Training	80.24	73.78	77.70	0.54	0.83
	Validation	73.44	71.43	72.64	0.44	0.78

These genes are specifically involved in cancer hallmark biological processes (Cancer hallmark GO terms). The model was evaluated using 10-fold cross validation on training dataset as well as on independent external validation dataset.