A Hybrid Feature Selection Method Based on Rough Conditional Mutual Information and Naive Bayesian Classifier

Table 2

Number and accuracy of selected features with different algorithms tested by naive Bayes.

Number	Unselect	CFS		Consistency		mRMR		RCMI
	Accuracy	Accuracy	Feature number	Accuracy	Feature number	Accuracy	Feature number	Accuracy	Feature number
1	75.00%	78.54%	22	75.66%	24	75.00%	21	77.43%	16
2	84.52%	89.03%	8	85.81%	13	87.74%	7	85.13%	8
3	90.60%	92.59%	13	90.88%	7	90.60%	7	94.87%	6
4	91.52%	93.51%	8	93.20%	9	93.98%	5	93.07%	3
5	85.58%	77.88%	19	82.21%	14	84.13%	10	86.06%	15
6	92.09%	91.80%	24	84.63%	13	90.48%	21	91.22%	24
7	90.11%	94.71%	5	91.95%	12	95.63%	1	95.63%	2
8	95.78%	96.66%	11	96.84%	8	95.78%	9	95.43%	4
9	98.31%	98.88%	10	99.44%	5	98.88%	8	98.88%	5
10	95.05%	95.05%	10	93.07%	5	94.06%	4	95.05%	10
Average	89.86%	90.87%	13	89.37%	11	90.63%	9.3	91.28%	9.3