Improved ML-Based Technique for Credit Card Scoring in Internet Financial Risk Control

Table 10

Comparison of accuracy rates yielded by different classifiers.

Model	Cutoff point	Training data			Test data
		Classification	Discriminant accuracy	Total accuracy	Classification	Discriminant accuracy (%)	Total accuracy (%)
Logistic regression	0.35	Nondefaults	72.9	71.5	Nondefaults	69.7	70.1
		Defaults	70.1		Defaults	70.5
GMDH	0.41	Nondefaults	85.1	79.4	Nondefaults	75.1	75.1
		Defaults	73.7		Defaults	75.1
SVM	0.26	Nondefaults	88.3	83.1	Nondefaults	78	77.4
		Defaults	77.9		Defaults	76.7
Proposed method	0.6	Nondefaults	100	100	Nondefaults	89.6	90.1
		Defaults	100		Defaults	90.6