A MOOCs Recommender System Based on User’s Knowledge Background

Abstract

Massive open online courses (MOOCs) are becoming a novel and flexible way for education. Learning on the MOOCs platform is a continuous process with multiple independent selections of courses. Meanwhile, students’ choices of subsequent courses are affected by their knowledge background, i.e., their historical courses and their learning performance on these courses. However, existing MOOC recommender systems often neglect students’ knowledge background. To fill this gap, in this paper we propose a knowledge-aware recommendation model which incorporates users’ historical enrolled courses and learning feedback. Specifically, we first construct a Course Knowledge Graph to represent the learning sequence between MOOCs and a Keyword Knowledge Graph to model the relation between course titles. With these two knowledge graphs and the course titles, we can extract the content feature and knowledge graph feature of courses. Second, we combine the extracted features with the user’s learning feedback to learn the representation of his knowledge background according to his historical enrollments. Third, to model the different impacts of historical courses, we use an attention module to calculate the weights of these courses and aggregate the user’s historical representation, and finally predict the probability of the user enrolling a course. We conduct a series of experiments compared with state-of-the-art baseline methods across multiple popular metrics. The promising results show that our proposed method can effectively recommend MOOCs to students.