Application of Discretization in the Use of Logistic Financial Rating

There are several limitations in the application of the logistic financial rating theory, for example, it requires the continuous financial data to follow a logistic distribution, and the model has a problem of multicollinearity. In order to solve these problems, this paper used a new method by preprocessing the financial data with discretization before fitting the logistic model. 780 Chinese corporate bonds and 72 financial indices were used in the empirical study. Results show that a reasonable logistic model with not only less explanatory variables but also high prediction accuracy could be established in this way. Conclusively, this algorithm can provide an effective way to solve the restrictions and produce a more concise and reliable financial rating model by introducing discretization.