Rule-Enhanced Neural Learning Path Recommendation for Personalized E-Learning

The rapid growth of e-learning systems has intensified the demand for personalized learning path recommendation methods that deliver tailored, coherent, and efficient educational experiences. Existing approaches, including traditional collaborative filtering, knowledge tracing models like Bayesian Knowledge Tracing, and reinforcement learning-based systems such as the DKVMN-LB + RL method, face limitations in modeling prerequisite relationships, capturing contextual richness, and achieving computational efficiency. This paper introduces the Rule-Enhanced Neural Learning Path Recommendation method, a novel framework that integrates high-confidence association rule mining with a deep neural network to address these gaps. REN-LPR employs the Apriori algorithm to mine prerequisite rules, ensuring structured learning paths with a coherence score of 96.8%, and leverages a comprehensive feature set—including student performance, question attributes, rule-derived prerequisite status, and temporal dynamics—to enhance prediction accuracy and knowledge mastery. Extensive experiments on the ASSISTment2009 and ASSISTment2015 datasets demonstrate REN-LPR’s superiority over the baseline DKVMN-LB + RL and collaborative filtering, achieving a learning efficiency of 0.0090 and processing time of approximately 150 seconds.