Applying back-propagation neural network for financial early warning in listed manufacturing companies

Since the integration of the world economy, competition among various enterprises has been intensified. In addition, the impact of the epidemic and trade policies in recent years has had a significant impact on Chinese enterprises, especially the manufacturing industry, which has gradually entered a downturn. This study proposes a financial early warning model using Backpropagation Neural Network (BPNN) to help manufacturing enterprises to develop stably. The model strengthens the correlation degree of nodes in the BPNN structure by increasing the correlation coefficient and gray correlation. The study selects 26 indicators based on the proposed index selection principle for the information of manufacturing enterprises. Finally, factor analysis is used to reduce redundant information in the early warning indicators to improve the operational efficiency of the model. In the experiment, during the validation period, the MSE value of tansig was 0.10, and the optimal MSE value of trainlm was 0.06. The overall average accuracy of the FEWM was 91.6%. The established warning model has a good effect and can timely detect whether the enterprise has a financial crisis.