A Novel Selective Ensemble Algorithm for Imbalanced Data Classification Based on Exploratory Undersampling

Table 8

Pairwise comparisons of all algorithms based on the 𝐺-mean criterion.

Algorithms		CART	RUSB	SMOB	UNBag	SMBag	AdaC	RAMO	RotF	Easy	RotE-un	RotEasy
Mean		0.7103	0.7765	0.7998	0.8320	0.7513	0.6168	0.7527	0.7069	0.8432	0.8382	0.8746
CART	𝑟̇		1.0917	1.1389	1.2038	1.0672	0.7524	1.0469	0.9351	1.2187	1.2108	1.2709
	𝑠		17-0-3	19-0-1	20-0-0	18-0-2	11-0-9	16-0-4	13-0-7	20-0-0	20-0-0	20-0-0
	𝑞		0.0015	0.0000	0.0000	0.0002	0.6636	0.0072	0.1892	0.0000	0.0000	0.0000
RUSB	𝑟̇			1.0432	1.1027	0.9775	0.6891	0.9590	0.8565	1.1163	1.1091	1.1641
	𝑠			13-0-7	13-0-7	3-0-17	4-0-16	7-0-13	4-0-16	17-0-3	16-0-4	20-0-0
	𝑞			0.1892	0.1892	0.0015	0.0072	0.1892	0.0072	0.0015	0.0072	0.0000
SMOB	𝑟̇				1.0570	0.9370	0.6606	0.9192	0.8210	1.0700	1.0631	1.1159
	𝑠				12-0-8	1-0-19	1-0-19	3-0-12	1-0-19	15-0-5	12-0-8	18-0-2
	𝑞				0.3833	0.0000	0.0000	0.0015	0.0000	0.0266	0.3833	0.0002
UNBag	𝑟̇					0.8865	0.6250	0.8697	0.7768	1.0124	1.0058	1.0557
	𝑠					2-0-11	4-0-16	5-0-15	4-0-16	15-0-5	13-0-7	20-0-0
	𝑞					0.0002	0.0072	0.0266	0.0072	0.0266	0.1892	0.0000
SMBag	𝑟̇						0.7050	0.9810	0.8762	1.1420	1.1346	1.1909
	𝑠						9-0-11	15-0-5	8-0-12	20-0-0	20-0-0	20-0-0
	𝑞						0.6636	0.0266	0.3833	0.0000	0.0000	0.0000
AdaC	𝑟̇							1.3915	1.2429	1.6198	1.6093	1.6892
	𝑠							16-0-4	12-0-8	20-0-0	20-0-0	19-0-1
	𝑞							0.0072	0.3833	0.0000	0.0000	0.0000
RAMO	𝑟̇								0.8932	1.1640	1.1565	1.2139
	𝑠								2-0-18	17-0-3	17-0-3	19-0-1
	𝑞								0.0002	0.0015	0.0015	0.0000
RotF	𝑟̇									1.3033	1.2949	1.3591
	𝑠									19-0-1	19-0-1	19-0-1
	𝑞									0.0000	0.0000	0.0000
Easy	𝑟̇										0.9935	1.0428
	𝑠										9-0-11	18-0-2
	𝑞										0.6636	0.0002
RotE-un	𝑟̇											1.0496
	𝑠											18-0-2
	𝑞											0.0002