The indicator variables used in the credit score model have an important effect on the performance of the model. The scientific standardization of the index selection method needs to be further improved. This paper studies the application of the feature selection method in the field of machine learning to quantitatively determine the credit score model. Taking the actual credit evaluation problem as an example, this paper compares the four methods of feature selection (ReliefF method, correlation method, consistency method and package). It is proved that the feature selection method can be used in streamlining, Accuracy of the three aspects of credit rating to improve the performance of the model. Which is based on the consistency of the method and package performance is superior to the Reli-efF method and the relevance-based approach.

Download