Optimizing learning paths: Course recommendations based on graph convolutional networks and learning styles

Abstract

With the rise of Massive Open Online Course (MOOC) platforms and the growing popularity of self-directed learning, an increasing number of learners are utilizing online platforms to access educational resources. While these extensive course resources offer learners diverse and accessible learning experiences, they also present challenges in personalized course selection. Traditional recommendation models often lack sufficient interpretability and fail to effectively leverage the interactive data generated during curriculum learning or account for the impact of individual learning styles on recommendations. To address these limitations, this study proposes a novel model, Course Recommendations based on Graph Convolutional Networks and Learning Styles to Optimize Learning Paths. Firstly, learner-course interaction data is recursively propagated through graph convolutional networks to generate predictive scores for courses. Secondly, a matching scale between courses and learning styles is established to compute similarity scores. Finally, the predictive scores and learning style similarity scores are integrated to achieve personalized course recommendations. The experimental results on the MOOCCube dataset demonstrate that CGCNLS significantly outperforms the baseline methods across multiple evaluation metrics, and the average performance of Precision, Recall and NDCG is improved by 6.94 %, 6.63 % and 7.98 %, respectively, under different Top-K Settings (K = 5, 10, 20, and 30), which can more effectively recommend courses for learners. The findings of this research provide robust support for further advancements in recommender systems and are expected to enhance the user experience and learning outcomes on online learning platforms.