Graph Theory for Enhanced Feedback and Student Performance Prediction in Higher Education Using a GAT-BiLSTM-Based RL Model

Predicting student performance and enhancing feedback systems in higher education are significant difficulties. Traditional machine learning models don’t properly represent complex student interactions, resulting in poor forecast accuracy. This study presents a Graph Attention Network (GAT) and a Bidirectional Long Short-Term Memory (BiLSTM)-based Reinforcement Learning (RL) model to improve the prediction of student performance. The system analyses student participation through graph-based feature extraction and sequential learning, enhancing feedback through reinforcement learning. The experimental findings indicate that the suggested model attains an accuracy of 94.6%, precision of 92.8%, and an $F 1$-score of 93.1%, substantially surpassing the performance of Logistic Regression, Decision Tree, and Random Forest models. This strategy delivers a 7% enhancement in memory and $F 1$-score relative to current studies, illustrating its efficacy in personalised teaching. The research highlights the promise of AI-driven adaptive learning, facilitating future progress in intelligent educational systems.