Course Recommendation on Online Learning Platforms using Collaborative Filtering and Content-based Filtering with Implicit Feedback

Non-formal education is possible to be obtained through online course platforms nowadays. Due to the vast number of courses in an online learning platform, a recommender system is needed as it helps to suggest a course that matches one's preference. A collaborative filtering type of recommender system is more suitable for non-formal education. Besides that, a recommender system needs feedback to give suggestions, which often rely on explicit feedback. However, most publicly available datasets consist only of implicit feedback. Therefore, we employ two collaborative filtering recommendation methods that can utilize implicit feedback, namely Bayesian Personalized Ranking (BPR) and Collaborative Less is More Filtering (CLiMF). A feature augmentation based on a content-based filtering technique is also performed to reduce the sparsity of the dataset. Using both MOOC Cube and Canvas Network datasets, the experiments show that BPR performs better than CLiMF on both datasets. On the other hand, the use of content-based filtering with feature augmentation does not significantly affect the performance.