The field of Multiple Criteria Decision Analysis (MCDA) deals with alternatives evaluated by several criteria, aiming to recommend the “best” decision to the decision-maker (DM). In this context, we are interested in the indirect learning paradigm which is comparable to machine learning tasks as it consists of inferring from past observations of the DM, the model parameters that suit the DM’s preferences. Our model (MR-Sort) stems from the MCDA family of outranking models, where an alternative a outranks another alternative b if there is a strong support of criteria (a majority in MRSort) that favors a compared to b. In the literature, methods and algorithms used for sorting problems - classification into predefined and ordered categories -...