Research on Data-Driven Student Personalized Learning Path Recommendation System

This paper first uses an adaptive graph convolutional network (AGCN) to model the relationship between students and learning resources. AGCN can extract potential relationship information from student behavior data, effectively capture students' learning needs and preferences, and make the recommendation results more accurate. Secondly, reinforcement learning is used to optimize the selection of learning paths, and dynamically adjust the recommendation strategy according to students' learning progress and feedback, so as to ensure that students' paths are optimized at different learning stages and avoid falling into inefficient learning modes. Through this method, the system can not only make adjustments according to students' real-time needs, but also continuously improve the learning path recommendation effect. This paper uses a real learning behavior dataset of an educational platform and compares it with traditional collaborative filtering methods and deep learning-based recommendation methods. The experimental results show that the algorithm improves the recommendation accuracy and recall rate by more than 15 % compared with traditional methods, and the students' learning progress and knowledge mastery are also significantly improved, indicating that the algorithm can effectively improve the effect of personalized learning path recommendation.