A PROMETHEE II AND WARPED K-MEANS APPROACH TO ORDERED CLUSTERING: APPLICATION TO THE GLOBAL HEALTH SECURITY INDEX

Several scholars have proposed clustering strategies to partition a set of alternatives into preference-ordered clusters, based on non-compensatory multi-criteria decision analysis (MCDA) methods. Thus, some approaches using outranking-based methods have been proposed to build a partitioning, finding representative centroids of clusters and minimizing \emph{ex-post} the preference inconsistencies among alternatives assigned to different groups. In this study, we propose a multi-criteria clustering approach which first rank the set of alternatives and then performs an iterative algorithm to find ordered clusters, their cluster centroids, and the upper and lower boundaries of each cluster. An application to the Global Health Security Index (GHSI) is performed, considering 195 countries. Results are compared with the GHSI segments, showing the main differences between these approaches. Limitations are discussed and future research is proposed.