In a context of Multi-Criteria Decision Aiding (MCDA) which typically involves multiple uncertainties or information imperfections, we address the problem of ordered clustering. This problem has gained increasing attention in the field of MCDA in the last decade, especially in light of the parallel growth of data-analysis techniques for clustering purposes. The proposed model is based on the multiple-criteria aggregation procedure for mixed evaluations (MCAP-mix), designed to deal with different types of information imperfections such as imprecision, uncertainty or ambiguity. The distances between preference relations generated for each pair of alternatives by MCAP-mix are used as a measure of closeness between the alternatives. Based on these distances, an extension of the K-medoids clustering algorithm is developed through an Integer Programming (IP) formulation to partition the set of alternatives into ordered clusters. The model is illustrated through a numerical example.
Bienvenue à tous!