Research on Credit Risk Prediction under Unbalanced Dataset Based on Ensemble Learning

Table 5

Comparison of classification effect under different b, ε

Algorithm	Evaluation index	b = 0.5, ε = 0.3	b = 0.5, ε = 0.5	b = 1, ε = 0.3	b = 1, ε = 0.5
LightGBM	F1-score	0.9715	0.9729	0.9765	0.9773
	-mean	0.6812	0.6883	0.6931	0.7226
	AUC	0.8251	0.8260	0.8309	0.8360
	KS	0.5150	0.5067	0.5194	0.5409
XGBoost	F1-score	0.9718	0.9707	0.9746	0.9765
	-mean	0.6645	0.6584	0.6744	0.6916
	AUC	0.8124	0.8174	0.8261	0.8268
	KS	0.5376	0.5185	0.5388	0.5401
GBDT	F1-score	0.9602	0.9710	0.9702	0.9705
	-mean	0.6307	0.6683	0.6499	0.6704
	AUC	0.7892	0.7901	0.7840	0.7903
	KS	0.4483	0.4597	0.4537	0.4618
RF	F1-score	0.9701	0.9612	0.9665	0.9709
	-mean	0.6572	0.6501	0.6487	0.6786
	AUC	0.7885	0.7861	0.7802	0.8077
	KS	0.4669	0.4727	0.4588	0.4883
LR	F1-score	0.8298	0.8315	0.8243	0.8334
	-mean	0.6310	0.6441	0.6528	0.6546
	AUC	0.7420	0.7627	0.7624	0.7645
	KS	0.3755	0.4156	0.4096	0.4133