Cost-Sensitive Radial Basis Function Neural Network Classifier for Software Defect Prediction

Table 8

Comparison results and error analysis on NASA datasets.

NASA datasets	Techniques	Sensitivity	Specificity	FPR or	Balance	Accuracy	AUC	MSE (error)
CM1	Naïve Bayes [5]	71.03	78.65	34.09	68.37	64.57	0.75	0.1456
	Random Forest [5]	70.09	71.29	32.17	68.94	60.98	0.74	0.2314
	C4.5 Miner [5]	74.91	74.66	27.68	73.58	66.71	0.53	0.3765
	Immunos [5]	73.65	75.02	30.99	71.24	66.03	0.63	0.1732
	ANN-ABC [5]	75.00	81.00	33.00	71.00	68.00	0.77	0.2435
	Hybrid self-organizing map [13]	70.12	78.96	30.65	69.73	72.37	0.80	0.0810
	Support vector machine [14]	78.97	79.08	31.27	73.35	78.69	0.79	0.0154
	Majority Vote [14]	79.80	80.00	30.46	74.16	77.01	0.81	0.1968
	AntMiner+ [14]	80.65	78.88	30.90	74.22	79.43	0.84	0.0345
	Proposed ADBBO-RBFNN model	81.92	80.96	29.71	75.41	82.57	0.90	0.0067
JM1	Naïve Bayes [5]	68.98	69.77	36.54	66.11	60.78	0.68	0.6547
	Random Forest [5]	66.72	72.38	33.47	66.62	63.97	0.75	0.6721
	C4.5 Miner [5]	69.08	68.55	40.67	63.87	62.35	0.61	0.5498
	Immunos [5]	70.99	70.21	43.00	63.32	64.55	0.63	0.4219
	ANN-ABC [5]	71.00	73.05	41.00	64.00	61.00	0.71	0.4057
	Hybrid self-organizing map [13]	71.02	74.90	40.57	64.75	72.33	0.82	0.5692
	Support vector machine [14]	70.89	79.00	39.87	65.09	70.32	0.81	0.3759
	Majority Vote [14]	74.65	73.46	40.36	66.30	75.92	0.83	0.0345
	AntMiner+ [14]	75.81	80.96	37.12	68.67	74.51	0.72	0.1786
	Proposed ADBBO-RBFNN model	79.85	82.31	36.22	70.69	77.03	0.87	0.0156
KC1	Naïve Bayes [5]	74.33	76.85	35.71	68.90	65.87	0.79	0.9854
	Random Forest [5]	72.54	75.89	37.91	66.90	67.99	0.80	0.6231
	C4.5 Miner [5]	76.42	75.64	34.05	70.71	68.01	0.64	0.7893
	Immunos [5]	78.05	72.91	36.92	69.63	63.55	0.71	0.6451
	ANN-ABC [5]	79.00	77.00	33.00	72.00	69.00	0.80	0.2257
	Hybrid self-organizing map [13]	80.92	80.94	35.67	71.40	78.43	0.86	0.1847
	Support vector machine [14]	81.37	81.27	28.96	75.65	79.24	0.83	0.5467
	Majority Vote [14]	82.65	85.62	30.98	74.89	79.66	0.85	0.0578
	AntMiner+ [14]	84.29	84.99	26.11	80.40	80.51	0.90	0.0346
	Proposed ADBBO-RBFNN model	85.67	87.95	20.24	82.46	84.96	0.92	0.0239
KC2	Naïve Bayes [5]	77.24	75.98	24.57	76.32	74.00	0.82	0.1453
	Random Forest [5]	70.32	70.71	23.90	73.05	77.81	0.82	0.5498
	C4.5 Miner [5]	69.87	74.67	29.19	70.34	76.54	0.67	0.6672
	Immunos [5]	76.51	75.92	25.06	75.71	72.90	0.73	0.4591
	ANN-ABC [5]	79.00	76.00	21.00	79.00	79.00	0.85	0.3195
	Hybrid self-organizing map [13]	80.98	77.82	23.09	78.85	85.98	0.91	0.1666
	Support vector machine [14]	84.35	78.96	25.61	78.78	87.12	0.88	0.2789
	Majority Vote [14]	86.71	84.77	20.38	82.80	83.47	0.82	0.1087
	AntMiner+ [14]	86.07	83.98	21.88	81.66	90.86	0.80	0.0985
	Proposed ADBBO-RBFNN model	87.96	86.24	17.93	84.73	95.65	0.95	0.0067
PC1	Naïve Bayes [5]	87.98	82.34	42.31	68.90	60.00	0.70	0.7689
	Random Forest [5]	82.31	80.99	46.71	64.68	63.98	0.85	0.6792
	C4.5 Miner [5]	76.58	81.76	38.24	68.29	62.18	0.68	0.5564
	Immunos [5]	81.99	79.66	39.00	69.62	61.73	0.64	0.4987
	ANN-ABC [5]	89.00	83.00	37.00	73.00	65.00	0.82	0.3125
	Hybrid self-organizing map [13]	86.79	85.67	35.60	73.15	95.87	0.87	0.1325
	Support vector machine (SVM) [14]	80.98	86.59	34.98	71.85	92.45	0.76	0.2037
	Majority Vote [14]	84.61	84.37	36.08	72.26	92.50	0.85	0.1078
	AntMiner+ [14]	89.34	87.12	37.29	72.58	91.85	0.91	0.0987
	Proposed ADBBO-RBFNN model	90.89	89.33	30.23	73.49	96.29	0.93	0.0379