Table 7.

Comparison of the performance of the proposed dynamic fairness re-weighting method (DFW) against three other effective methods applied to the XGBoost model: Grid Search (GS), Exponentiated Gradient Reduction (EGR), and Correlation Remover (CR) on the test dataset. Additionally, we compare the performance of DFW with our baseline model, which incorporates Demographic Parity Loss (DPL). Each row: specifies the definition of the sensitive attribute; Each column: specifies the fairness metric to achieve. The disparity is defined as the largest difference in the fairness metric among different subgroups. A smaller disparity is better (bold).

	GS	EGR	CR	DPL	DFW	GS	EGR	CR	DPL	DFW	GS	EGR	CR	DPL	DFW
	AUROC					Recall						Negative predictive value (NPV)
Ethnicity
Hispanic$(\text{N}=355)$	0.8869	0.8264	0.8856	0.8424	0.8347	0.6604	0.6415	0.6604	0.8289	0.7632	0.9104	0.9073	0.9122	0.9350	0.9127
Non-hispanic$(\text{N}=\mathrm{1,693})$	0.9125	0.8287	0.9120	0.8607	0.8345	0.6585	0.6690	0.6655	0.8294	0.7640	0.8929	0.8948	0.8943	0.9211	0.9151
Disparity(Max-Min)	0.0256	0.0023	0.0264	0.0183	0.0002	0.0019	0.0275	0.0051	0.0005	0.0008	0.0175	0.0125	0.0179	0.0139	0.0024
Gender
Female$(\text{N}=926)$	0.9205	0.8592	0.9259	0.8549	0.8109	0.6987	0.6923	0.6859	0.7652	0.7174	0.9056	0.9028	0.9026	0.9261	0.9092
Male$(\text{N}=\mathrm{1,122})$	0.9063	0.8205	0.8925	0.8396	0.8172	0.6087	0.6413	0.6467	0.7956	0.7336	0.8841	0.8920	0.8934	0.9179	0.9050
Disparity(Max-Min)	0.0142	0.0387	0.0334	0.0153	0.0063	0.0900	0.0510	0.0392	0.0304	0.0162	0.0215	0.0108	0.0092	0.0082	0.0042
Race
Asian$(\text{N}=36)$	0.9394	0.9167	0.9646	0.9722	0.8667	0.7778	0.8889	0.8889	0.7889	0.8333	0.9130	0.9545	0.9524	0.9200	0.9310
White$(\text{N}=599)$	0.8717	0.8301	0.8905	0.7905	0.8074	0.5532	0.5957	0.6277	0.6547	0.7589	0.8783	0.8869	0.8955	0.8490	0.9326
African$(\text{N}=309)$	0.8918	0.8280	0.9169	0.7935	0.7856	0.5000	0.6071	0.6964	0.7194	0.8023	0.8519	0.9091	0.9012	0.8554	0.8970
Other$(\text{N}=\mathrm{1,102})$	0.8889	0.8388	0.9015	0.8502	0.8518	0.5680	0.6748	0.6553	0.6878	0.8155	0.8455	0.8797	0.8737	0.8571	0.9314
Disparity(Max-Min)	0.0677	0.0887	0.0741	0.1817	0.0811	0.2778	0.2932	0.2612	0.1342	0.0744	0.0675	0.0748	0.0787	0.0710	0.0356