Boosting methods for multi-class imbalanced data classification: an experimental review

Journal of Big Data

Table 5 Ranks computed by the Wilcoxon test based on the MAUCs

Algorithms	AdaBoost.MH	SAMME	CatBoost	LogitBoost	GradientBoost	XGBoost	MEBoost	SMOTEBoost	RUSBoost	LightGBM	AdaC1	AdaC2	AdaC3	AdaCost
AdaBoost.MH	−	61	20.5	39.5	64	26	85	36	73	85	74	75	64	55
SAMME	44	−	14	31	57	35	65	23	76	75	86	83	62	57
CatBoost	99.5	106	−	76	95	79	106	79	109	101	107.5	107	95	90
LogitBoost	80.5	89	44	−	81	46	73	64	83	89	102	98	88	83
GradientBoost	56	63	25	39	−	45	64	43	75	62	76	64	73	53
XGBoost	94	85	41	74	75	−	87	58	81	85	105	96	80.05	80
MEBoost	35	55	14	32	56	33	−	32	62	85	72	60	52	51
SMOTEBoost	84	97	41	56	77	62	88	−	102	88	109	111	94	88
RUSBoost	47	44	11	37	45	39	58	18	−	68	70	64	43	47
LightGBM	35	45	19	31	58	35	35	32	52	−	58	43	42	41
AdaC1	46	34	12.5	18	44	15	48	11	50	62	−	43.5	41	25
AdaC2	45	37	13	22	65	24	60	9	56	77	76.5	−	43	40
AdaC3	56	58	25	32	47	39	68	26	77	78	79	77	−	52
AdaCost	65	63	30	37	67	40	69	32	73	79	80	80	68	−