A ResNet-LSTM Based Credit Scoring Approach for Imbalanced Data

Table 3

Comparison of different data generation algorithm (Bank dataset).

	Critics	None	ADASYN	SMOTE	BorderlineSMOTE	CGAN	ACTGAN
Support vector machine	Recall	97.53	99.18	99.07	99.25	99.13	99.02
	F1	77.01	90.66	89.66	89.75	89.37	89.27
	AUC	94.52	95.38	95.11	95.21	95.56	96.02
	KS	88.41	84.59	82.90	83.25	83.48	89.01
Logistic regression	Recall	79.43	97.95	98.75	98.68	89.54	92.60
	F1	72.28	90.31	89.68	89.79	80.92	81.84
	AUC	86.50	94.91	95.05	95.13	90.37	92.40
	KS	70.01	82.98	82.46	82.72	84.65	85.54
Decision tree	Recall	62.99	86.59	85.12	86.11	69.24	89.24
	F1	63.20	87.23	85.17	85.42	71.05	81.05
	AUC	78.73	90.74	89.58	89.93	81.65	91.34
	KS	53.53	81.04	77.71	78.93	61.57	81.69
K nearest neighbors	Recall	24.70	90.22	88.98	89.48	94.46	95.76
	F1	13.45	88.26	87.11	87.33	84.95	84.05
	AUC	16.06	92.06	91.40	91.63	94.42	94.09
	KS	10.04	80.12	82.78	82.69	86.17	87.46