Prediction of Students' Performance in E-learning Environments Based on Link Prediction in a Knowledge Graph

In recent years, the growing need for easily accessible high-uality educational resources, supported by the advances in AI and Web technologies, has stimulated the development of increasingly intelligent learning environments. One of the main requirements of these smart tutoring systems is the capacity to trace the knowledge acquired by users over time, and assess their ability to face a specific Knowledge Component in the future with the final goal of presenting learners with the most suitable educational content. In this paper, we propose a model to predict students' performance based on the description of the whole learning ecosystem, in the form of a RDF Knowledge Graph. Subsequently, we reformulate the Knowledge Tracing task as a Link Prediction problem on such a Knowledge Graph and we predict students outcome to questions by determining the most probable link between each answer and its correct or wrong realizations. Our first experiments on a real-world dataset show that the proposed approach yields promising results comparable with state-of-the-art models.