How to balance the bioinformatics data: pseudo-negative sampling

BMC Bioinformatics

Table 3 Performance comparison of classifiers under different percentage of pseudo-negative samples on the Haberman data

Percentage	Classifier	Sen(%)	Spe(%)	Acc(%)	MCC
0	DA	17.33	95.42	74.79	0.212
	AdaBoost	29.19	90.89	74.71	0.266
	RF	34.2	82.84	70.07	0.197
	NN	27.98	87.28	71.68	0.202
10	DA	21.77	93.96	72.96	0.236
	AdaBoost	32.72	86.12	70.58	0.214
	RF	33.38	83.91	69.38	0.197
	NN	30.37	82.01	67.04	0.144
20	DA	30.51	94.41	74.2	0.340
	AdaBoost	46.68	87.54	74.26	0.370
	RF	45.01	81.32	69.59	0.272
	NN	37.42	82.97	68.57	0.222
30	DA	31.73	95.1	73.32	0.36
	AdaBoost	51.81	87.15	75.65	0.422
	RF	51.06	79.6	70	0.311
	NN	42.39	84.54	70.36	0.291
40	DA	37.13	94.38	72.93	0.404
	AdaBoost	50.73	86.1	72.87	0.396
	RF	56.81	78.38	69.95	0.359
	NN	53.63	81	70.6	0.35
50	DA	38.61	93.83	71.74	0.405
	AdaBoost	61.46	82.26	73.81	0.447
	RF	60.75	78.22	70.95	0.395
	NN	52.41	79.81	68.56	0.339

ISSN: 1471-2105