Preference-based clustering of large datasets

International audience
Clustering has been widely studied in Data Mining literature, where, through different measures related to similarity among objects, potential structures that exist in the data are uncovered. In the field of Multiple Criteria Decision Analysis (MCDA), this topic has received less attention, although the objects in this case, called alternatives, relate to each other through measures of preference, which give the possibility of structuring them in more diverse ways. In this paper we present an approach for clustering sets of alternatives using preferential information from a decision-maker. As clustering is dependent on the relations between the alternatives, clustering large datasets quickly becomes impractical, an issue we try to address by extending our approach accordingly.