Software Defect Prediction through Neural Network and Feature Selections

Table 5

RBF training and testing results using selected features in the data set.

Data set	Selected feature number	Training results				Testing results
		Precision	Recall	F-measure	Accuracy	Precision	Recall	F-measure	Accuracy
CM1	7	97.10	97.77	0.99	95.30	94.01	95.78	0.95	93.99
JM1	8	88.87	90.09	0.90	87.83	85.18	87.89	0.87	84.87
KC1	8	85.78	86.89	0.86	84.19	83.24	86.24	0.83	83.25
KC2	3	83.19	84.67	0.84	82.88	81.27	83.82	0.82	79.11
KC3	4	84.99	85.39	0.86	84.19	79.30	82.10	0.85	78.25
KC4	5	84.78	85.99	0.89	83.89	85.29	86.28	0.86	83.18
MC1	10	99.79	100	1.00	100	99.89	100	0.99	99.01
MC2	8	72.27	73.75	0.79	71.72	73.27	76.67	0.76	70.18
MW1	11	91.28	94.16	0.99	90.71	90.90	92.09	0.95	88.90
PC1	6	100	100	1.00	100	98.79	99.98	0.99	98.99
PC2	9	100	100	1.00	100	100	100	0.99	99.80
PC3	6	95.39	96.37	0.99	95.38	95.67	96.23	0.97	94.11
PC4	8	97.96	98.94	0.98	96.49	95.12	95.17	0.95	94.44
PC5	5	83.67	85.96	0.84	81.89	80.89	81.80	0.80	79.02