TCR: topologically consistent reweighting for XGBoost in regression tasks

Table 1 Dataset overview, where N, UT, S, UT/S, NF:CF, UF, R and ID indicate the name, the number of unique targets, the set size, the quotient of the previous two, the numbers of nominal and continuous features (excluding the target), the number of unique feature values, the reference and the OpenML ID, respectively; “sk” indicates that the dataset is available on scikit-learn; the IDs in parentheses refer to the original sets

N	UT	S	UT/S	NF:CF	UF	R	\({\textrm{ID}}^{1}\)
Diabetes	214	442	0.48	0:10	1135	Efron et al. (2004)	\({\textrm{sk}}^{2}\)
Treasury	570	1049	0.54	0:15	8180	–/\({-}^{3}\)	42,367
Wine Quality -REG	7	6497	0.001	0:11	2119	Cortez et al. (2009)	287
Topo21	1336	8885	0.15	0:266	350,916	Feng et al. (2003)	422
Bike Sharing Demand	869	17,379	0.05	28:6	230	Fanaee-T and Gama (2014)	42,713
Mimic Los	16,420	19,196	0.86	12:31	64,254	(-by us-)	–/–
California Housing	3842	20,640	0.19	0:8	70,057	Kelley Pace and Barry (1997)	sk[2]
Online News Popularity	1454	39,644	0.04	14:44	457,216	Fernandes (2015)	4545
Diamonds	11,602	53,940	0.26	0:9	1137	–/–	42,225
BNG	78,571	78,732	0.998	4:7	543,871	–/–	1213
Superconduct	3007	21,263	0.14	0:79	573,105	Grinsztajn et al. (2022)	44,148 (43,174)
House Sales	4027	21,613	0.19	2:15	19,787	Grinsztajn et al. (2022)	44,066 (42,731)
Miami Housing 2016	13,932	13,932	1.0	0:13	108,276	Grinsztajn et al. (2022)	44,147 (43,093)
Mercedes Benz Gr. Manuf	2545	4209	0.60	376:0	2	Grinsztajn et al. (2022)	44,061 (42,570)
Yprop41 (\(\sim\) Topo21)	1336	8885	0.15	20:42	18,896	Grinsztajn et al. (2022)	44,054 (416)
Elevators	61	16,599	0.004	0:16	1945	Grinsztajn et al. (2022)	44,134 (216)
Isolet	26	7797	0.003	0:613	10,001	Grinsztajn et al. (2022)	44,135 (300)
CPU Act	56	8192	0.007	0:21	33,328	Grinsztajn et al. (2022)	44,132 (197)
Pol	11	15,000	0.001	0:26	200	Grinsztajn et al. (2022)	44,133 (201)
Ailerons	35	13,750	0.003	0:33	1933	Grinsztajn et al. (2022)	44,137 (296)