The paper proposes a two-phase methodology to support groups in multicriteria classification problems. The first phase, which relies on a dominance-based rough set approach (DRSA), takes a set of assignment examples as input and outputs a set of collective decision rules, representing a generalized description of the decision makers' preference information. The second phase then applies these collective decision rules to classify all decision objects. The methodology uses “if …then…”aggregation rules that coherently implement the majority principle and veto effect. The aggregation rules thus allow obtaining consensual decisions. Furthermore, the contribution of each decision maker to the collective decision is objectively measured by the quality of individual classification conducted by this decision maker during the first phase. The methodology has been validated by developing a prototype and applied to a nuclear risk management decision problem.