An Implicit Feedback Recommendation System for Massive Open Online Courses

Massive open online courses (MOOCs) have recently becoming a popular means of education. They generally give the students large-scale options. However, the diversity of MOOC courses available and their rapid updates make it more difficult for students to find fresh material relevant to them. A recommendation system (RS) connects the learner with the best learning resources to meet students' interests. The majority of recommender system research is based on the existence of explicit feedback, which is often impossible or inaccessible in MOOCs. As a result, in this paper, we model user positive and negative preferences using implicit feedback acquired passively by watching various types of students' behavior. This paper proposes a novel course recommendation, which employs Siamese Neural Networks (SNNs) to extract latent representations of students and courses using a loss function that favors observed over unobserved courses. The similarity of users and courses is then determined using a novel representation mechansim. Furthermore, recommending those courses with limited interaction data is a major challenge in MOOC recommenders. To tackle this issue, the courses profiles are used as side information which helps us create more accurate representations. To evaluate the performance of the propsoed method, we performed the experiments on a real dataset gathered from XuetangX—one of China's largest MOOCs. The results of the experiments show that the proposed method outperforms a number of baseline nad state-of-the-art MOOC recommenders.