Research on Enterprise Financial Risk Warning Based on AutoEncoder-ConvNeXt

Abstract

This study combines the AutoEncoder and ConvNeXt models to conduct financial risk warning research on 167 Chinese agro-forestry related enterprises. Firstly, a set of 52 indicators was determined by similarity calculation to further discuss the classification method of financial risk. The CRITIC method, combined with grey correlation analysis, fuzzy comprehensive evaluation, and TOPSIS method, was used to score the financial risk of agro-forestry enterprises. Based on this, a self-organizing mapping network was used to classify the financial risk level, and finally, the AutoEncoder-ConvNeXt model was used to predict the financial risk of enterprises. Compared with other models such as ResNet50 and original ConvNeXt, the prediction accuracy of the AutoEncoder-ConvNeXt model was higher at 87.11%, making it better suited for predicting the financial risks of listed companies.