MOOCRec: An Attention Meta-path Based Model for Top-K Recommendation in MOOC.

With the surge of the courses and users on Massive Open Online Courses (MOOC), MOOC has accumulated rich educational data. However, the utilization of MOOC resources is not high enough to satisfy the dynamic and diverse demands of different individuals. Meanwhile, the traditional recommendation model for MOOC dataset underperforms in both precision and recall. To address those issues, we collect and collate a MOOC dataset and then propose an attention meta-path based recommendation model named MOOCRec to jointly learn explicit and implicit relationships between students and courses. By extracting the knowledge points of the whole course information, we successfully construct different heterogeneous information networks (HINs) in MOOC and then we elaborately design multiple meta-paths based context to exploit the heterogeneity of other HINs in MOOC, which enables MOOCRec to offer abundant course resources. In particular, we leverage three attention mechanisms under MOOC to further enhance factors that effectively influence student preferences to improve the precision of our model. What’s more, we adopt another classical dataset called Movielens, reconstruct HINs and redesign meta-paths to demonstrate that the extensive availability of MOOCRec.