Recommendation Filtering à la carte for Intelligent Tutoring Systems

In computerized adaptive testing, the activities have to be well adjusted to the latent knowledge of the students. Collaborative and content-based filters are usually considered as two solutions of data-centric approach using the evaluation data to uncover the student abilities. Nevertheless, past lecturer recommendations can induced possible bias by using a single and immutable training set. We try to reduce this issue by releasing a hybrid recommendation filtering. Our approach is supported by the Item Response Theory and techniques of clustering to output purely objective recommendation filters selecting activities and building an evaluation path based on historical evolutions of past students. In this paper, we particularly highlight the crucial clustering task by offering plots and metrics to adjust the decisions of the practitioners.