A ResNet-LSTM Based Credit Scoring Approach for Imbalanced Data

Table 4

Comparison of different data generation algorithm (UCI dataset).

	Critics	None	ADASYN	SMOTE	BorderlineSMOTE	CGAN	ACTGAN
Support vector machine	Recall	22.70	23.54	40.99	42.13	46.07	47.32
	F1	34.15	35.77	50.92	51.96	55.72	56.98
	AUC	59.92	60.36	66.47	67.04	69.16	69.79
	KS	19.70	20.09	32.12	33.37	37.75	39.16
Logistic regression	Recall	25.05	27.49	33.84	35.83	39.67	43.13
	F1	36.41	39.71	46.10	47.95	52.05	55.17
	AUC	60.78	61.83	64.31	65.18	67.27	68.84
	KS	21.52	23.67	26.99	28.83	32.57	36.18
Decision tree	Recall	41.68	51.73	56.17	55.95	55.63	56.56
	F1	39.92	49.75	55.19	54.93	55.28	55.19
	AUC	61.50	66.04	68.56	68.38	68.64	68.31
	KS	22.05	31.25	36.52	36.95	36.60	36.04
K nearest neighbors	Recall	34.08	41.14	46.46	48.84	44.65	46.49
	F1	41.79	47.99	53.83	56.01	54.64	55.84
	AUC	63.00	65.28	67.97	69.28	68.55	69.12
	KS	24.00	29.74	34.12	36.46	36.51	37.89