Dominance-based rough set approach for group decisions

•The approach is designed to handle multiple decision makers in a way that leads to one compromise solution.•The preferences of decision makers are extracted using a learning set-based approach.•The judgments of decision makers are combined via a mixed input–output aggregation strategy.•The weights of decision makers are objectively computed based on the input data.•The assignment of decision objects to decision classes is based on the use of the majority principle and veto effect.•The majority principle and veto effect notions are implemented through the concepts of concordance and discordance.

The objective of this paper is to propose an approach to support group multicriteria classification. The approach is composed of three phases. The first phase exploits the knowledge provided by each decision maker to individually approximate the decision classes using rough approximation. The second phase seeks to combine the outputs of individual approximation phase into a collective decision table by using an appropriate aggregation procedure. The third phase uses the collective decision table in order to infer a set of collective decision rules, which synthesize the judgements and perspectives of the different decision makers and to permit the classification of all decision objects. The proposed approach relies on the Dominance-based Rough Set Approach (DRSA), which is used at two different levels. First, the DRSA is used during the first phase to approximate the input data relative to each decision maker. Second, the DRSA is used during the third phase to approximate the collective decision table and generate the collective decision rules. This paper presents the theoretical foundation of the proposed approach, three case studies using real-world data and a comparative study of recent similar proposals.