A MULTI-LAYER ARCHITECTURE FOR AN E-LEARNING HYBRID RECOMMENDER SYSTEM

Distance computer-assisted learning is increasingly common, owing largely to the expansion and development of e-technology. Nevertheless, the available tools of the learning platforms have demonstrated their limits during the pandemic context, since many students, who were used to ”face-to-face” education, got discouraged and dropped out of school. In this context, a main issue is to conceive tools that would allow the teachers to supervise their students at a distance, by monitoring their progress and ensuring follow-up action as required. Another issue is to equip the learning platforms with intelligent systems able to guide the students involved in pedagogic activities. In this work, we propose a novel architecture of recommender system for vocational higher education that provides the students with personalized advice and the teacher with suitable information, in order to make the task of monitoring easier and involving them in the machine learning. Our system is supposed to act in a hybrid environment, and, for this purpose, has to explain its predictions in an interpretable and faithful manner, both to the students and the teachers, so that the former can determine the relevance of what is suggested and the latest can act on the future analyses and recommendations. This is a multi-layer architecture, so that each step of the recommendation process is meaningful, thus explicable to the users. The design of this architecture is a preliminary stage of a recommender system. It is designed on top of a learning digital infrastructure exploited since 2018 by the 1000 students of Bayonne Institute of Technology.