Clues from networks: quantifying relational risk for credit risk evaluation of SMEs

Financial Innovation

Table 28 Discrimination performance of different feature sets under 90% confidence intervals

Model	Metric	BF	BF + RS	BF + CN	BF + DN	BF + SN
LR	AUC	0.715 ± 0.012	0.726 ± 0.012	0.722 ± 0.012	0.729 ± 0.012	0.725 ± 0.012
	KS	0.408 ± 0.018	0.419 ± 0.018	0.414 ± 0.018	0.431 ± 0.019	0.417 ± 0.018
	H	0.251 ± 0.017	0.270 ± 0.017	0.266 ± 0.017	0.283 ± 0.017	0.269 ± 0.016
RF	AUC	0.725 ± 0.012	0.738 ± 0.011	0.738 ± 0.011	0.741 ± 0.011	0.735 ± 0.011
	KS	0.426 ± 0.019	0.445 ± 0.017	0.444 ± 0.017	0.450 ± 0.018	0.439 ± 0.018
	H	0.265 ± 0.017	0.279 ± 0.015	0.289 ± 0.015	0.294 ± 0.016	0.281 ± 0.016
XGB	AUC	0.695 ± 0.011	0.715 ± 0.011	0.718 ± 0.011	0.722 ± 0.010	0.720 ± 0.012
	KS	0.387 ± 0.018	0.416 ± 0.018	0.415 ± 0.017	0.424 ± 0.015	0.434 ± 0.018
	H	0.219 ± 0.015	0.252 ± 0.016	0.253 ± 0.016	0.245 ± 0.014	0.257 ± 0.017

Each metric are measured in mean of 100 estimates obtained by repeating the outer ten-fold cross-validation procedure ten times (mean ± standard deviation from a 90% confidence interval)