Research on Enterprise Financial Risk Based on Radial Basis Function Neural Network

This article takes A-share manufacturing listed companies in the Shanghai and Shenzhen stock markets as the research subjects. Firstly, 22 financial indicators were preliminarily determined based on the selection principles of warning indicators. Then, financial data of companies in financial crisis and normal companies in 2023 were obtained from the Guotai An database, and descriptive statistics were conducted using SPSS software. Subsequently, data from 135 financially distressed companies and 270 financially normal companies were extracted and subjected to significant difference testing using STATA software. Ultimately, 19 financial indicators were determined, and two financial warning models were constructed. The following conclusions were drawn: (1) Manufacturing listed companies that have fallen into financial crisis difficulties have not been resolved, and the number of financially distressed companies each year remains basically unchanged. (2) Compared with financially normal companies, companies in financial crisis face greater debt repayment pressure, limited development space, weaker profitability, and lower operational efficiency; (3) The prediction performance of the neural network financial warning model and the multivariate discriminant financial warning model is good, with prediction accuracies of 86.8% and 84.8%, respectively, both reaching over 80%. Among them, the radial basis function neural network financial warning model has better predictive performance and can effectively control and prevent financial risks.