The identification of communities in graphs has been largely addressed in the literature, although the case of dynamic graphs still remains open. This article proposes an approach to compute representative communities in dynamic graphs, measure their stability and level of activity, and visualize their evolution over time.