KPCR: Knowledge Graph Enhanced Personalized Course Recommendation

To handle the limitations of collaborative filtering-based recommender systems, knowledge graphs are getting attention as side information. However, there are several problems to apply the existing KG-based methods to the course recommendations of MOOCs. We propose KPCR, a framework for Knowledge graph enhanced Personalized Course Recommendation. In KPCR, internal information of MOOCs and an external knowledge base are integrated through user and course related keywords. In addition, we add the level embedding module that predicts the level of students and courses. Through the experiments with the real-world datasets, we demonstrate that our knowledge graph boosts recommendation performance as side information. The results also show that the two auxiliary modules improve the recommendation performance. In addition, we evaluate the effectiveness of KPCR through the satisfaction survey of users of the real-world MOOCs platform.