<p>With the complexity of the global economic environment, enterprises are facing an increasing risk of financial distress. However, traditional financial risk assessment methods rely on linear assumptions, which have significant limitations in dealing with complex, high-dimensional, and nonlinear financial data of modern enterprises, making it difficult to accurately identify potential financial distress. To solve this problem, the study proposes a corporate financial distress prediction model based on graph convolutional neural networks and dynamic time regularization. The model firstly transforms the corporate financial data into graph structure, and extracts the features of complex financial relationships through graph convolutional neural network, and at the same time combines with the dynamic time regularization method to enhance the adaptability to the dynamic change of time. The experimental data are obtained from the China Stock Market and Accounting Research Database, covering the quarterly financial data of listed companies during the period from 2013 to 2020, which contains 120 financially distressed companies and 1,929 normal operating companies. The experimental results show that the proposed model achieves an accuracy of 67.47% in predicting financial distress, a recall rate of 72.36%, an F1 value of 68.58%, and a misclassification rate of less than 4%, which are all superior to traditional methods such as K-nearest neighbor, support vector machine and convolutional neural network. In addition, combined with business intelligence technology, a visualized financial forecasting system is constructed, which enhances the interpretability of the model and provides intelligent financial decision support for enterprise managers. The research results can provide theoretical basis for enterprise managers to optimize financial structure, investors to assess enterprise health, and the government to formulate financial regulatory policies.</p>