How to balance the bioinformatics data: pseudo-negative sampling

BMC Bioinformatics

Table 4 Performance comparison between MMPCC and SMOTE under different percentage of pseudo-negative samples

Percentage(%)	Methods	Classifiers	Sen(%)	Spe(%)	Acc(%)	MCC
10	MMPCC	RF	17.13	99.44	92.46	0.333
		NN	17.05	99.23	92.25	0.312
	SMOTE	RF	16.01	98.27	91.34	0.235
		NN	5.2	99.69	91.68	0.16
20	MMPCC	RF	17.28	99.45	91.84	0.337
		NN	24.6	99.22	92.31	0.405
	SMOTE	RF	17.07	98.14	90.75	0.246
		NN	8.05	99.49	91.16	0.2
30	MMPCC	RF	18.18	99.44	91.29	0.351
		NN	30.38	99.16	92.27	0.464
	SMOTE	RF	17.69	97.95	90.08	0.25
		NN	10.16	99.23	90.5	0.216
40	MMPCC	RF	19.09	99.43	90.75	0.363
		NN	35.94	99.08	92.26	0.513
	SMOTE	RF	18.54	97.8	89.5	0.258
		NN	12.07	99.14	90.02	0.243
50	MMPCC	RF	19.56	99.39	90.16	0.367
		NN	38.82	99.13	92.15	0.543
	SMOTE	RF	18.5	97.72	88.9	0.258
		NN	14.05	99.01	89.55	0.266

ISSN: 1471-2105