Link Prediction in Educational Graph Data to Predict Elective Course using Graph Convolutional Network Model

Abstract

Personalized learning has achieved the attention of many researchers in the Education field. Personalized learning is a teaching model in which students (learners) have a central role in the learning process. By using this approach, educational methods, and techniques are customized and adapted to be better suited for each learner, with their unique learning style, background, needs, and previous experiences. Based on what the learners have already learned, subjects have already known, and skills have already developed each student in a personalized learning process will get a "learning plan". This approach is different from a conventional approach or known as the "one size fits all" approach. The challenge of personalized learning is how to connect a learner’s previous knowledge, skills and with learning materials that will link that understanding with new knowledge. This paper presents a novelty technique to implement personalized learning by automating a predicted linkage between a student in higher education and elective courses based on previous learning achievement. In this study, Graph Convolutional Networks (GCNs) are used to address link prediction tasks between student and elective courses. The empirical results showed that the GCN model can be used to predict elective courses for a student with 62.5 % average accuracy.