Scholarlink

<p>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.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on Financial Risk Early Warning of Higher Education Institutions Using LSTM-CNN Enhanced by Attention Mechanism

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月23日

查看全文
  • Yu Chao,
  • Nur Fazidah Elias,
  • Yazrina Yahya,
  • Ruzzakiah Jenal

摘要

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.