Graph path fusion and reinforcement reasoning for recommendation in MOOCs

In recent years, online learning methods have gradually been accepted by more and more people. A large number of online teaching courses and other resources (MOOCs) have also followed. To attract students' interest in learning, many scholars have built recommendation systems for MOOCs. However, students need a variety of different learning resources, such as courses, videos, concepts, etc., and it is difficult to find suitable resources by themselves. So we propose a resource recommendation method called Multi-path Embedding and User-centric Reasoning (MEUR), which embeds multiple paths and searches with users as the center, innovatively combining the advantages of graph convolution network and reinforcement learning, ultimately shows the path of the knowledge graph. First, we put forward the problem to solve, which is to recommend multiple types of learning resources for users at the same time and show the corresponding reasoning path as the reason for the recommendation. Second, we propose an embedding model that integrates multi-path and graph convolution network, embedding entities in the knowledge graph into vectors. Third, we use reinforcement learning and combine user-centric reasoning to make recommendations for users. Finally, we use datasets from a real MOOC platform to evaluate our model through experiments and compare it with other methods.