Working in groups aims to exceed the result that could be obtained by the simple sum of results achieved individually. To this end, the definition of the group is a key element: How to choose the members of the latter, on what criteria to identify them? In our work we focus on the process of forming groups of learners taking into account the characteristics of learners and more generally the context in which they evolve. We specify what we mean by context before presenting our context-based decision support to form relevant groups of learners. Using a motivating example, we illustrate how to use the multi-criteria classification process for decision support. We then discuss its applicability within a collaboration platform developed from a semantic model.