Towards multicriteria analysis: A new clustering approach

Abstract

The researches in the multicriteria classification fields focus on the assignment of objects into predefined classes. Nevertheless, the construction of multicriteria clusters is not enough studied in the field of research. To deal with this problem, we propose a new clustering approach based on the definition of a new distance which takes into account the multicriteria nature of the problem. This distance uses the preference relations of the Promethee method and the Sokal and Michener index so widely used in the classification field. The approach generates, according to the preference relations 4 clustering. Each clustering expresses a way of grouping objects according to a preference relation. To get the final optimal clustering, an aggregation procedure, based on the minimization of the disagreements between the four clustering, is used.