Under increasing internal and external pressures, higher education institutions (HEIs) are grappling with escalating financial risks, necessitating the development of robust predictive frameworks. This study introduces an advanced hybrid model that combines Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN), augmented by an enhanced attention mechanism, to significantly improve the accuracy of financial risk forecasting. The study designs a comprehensive financial risk matrix that incorporates time-series analysis, grid structures, and other critical dimensions to provide a multidimensional approach to risk evaluation. The study systematically compares and analyzes the prediction curves of LSTM, CNN, LSTM-CNN, and LSTM-Attention-CNN models, highlighting the superior performance of the LSTM-Attention-CNN model. Additionally, this model is benchmarked against traditional approaches, including ARIMA, Logistic Regression, Support Vector Machines (SVM), and Random Forests, achieving marked improvements across multiple evaluation metrics. These results underscore the efficacy of the LSTM-Attention-CNN model in delivering early warnings and reliable decision support for HEIs. By integrating advanced deep learning methodologies with innovative matrix design, this research makes a substantive contribution to financial risk management. Future endeavors will further optimize the model and extend its applicability to dynamic, complex financial scenarios.
Chao et al. (Sat,) studied this question.