Research on corporate financial early warning model combining multiple regression and deep learning

Abstract

The global economy is complex and ever-changing, and corporate financial early warning has become the focus of attention in academic and practical circles. Traditional financial early warning models are primarily based on statistical methods, such as multiple regression analysis, which has limitations in dealing with high-dimensional and nonlinear data. This study combines various regression analysis and deep learning to build a more efficient and accurate corporate financial early warning model. Through deep learning to mine complex patterns of financial data and integrate them with multiple regression analysis results to improve the accuracy and timeliness of early warning. The financial data of 200 listed companies in the past 5 years are selected as samples, including 100 financially healthy enterprises and 100 financial crisis enterprises. First, multiple regression analysis is used to identify the key indicators that affect the financial status of enterprises, such as the asset-liability ratio and current ratio. Then, build a deep learning model, input the original financial data, and automatically extract features and classify and predict them through a multi-layer neural network. The experimental results show that the early warning accuracy of the single multiple regression model is 75%, and the accuracy of the hybrid model combined with deep learning is increased to 85%, significantly improving the prediction ability.