Clustering alternatives in the preference-approval context

Abstract

Preference-approval structures combine preference rankings and approval voting to express preferences over a set of alternatives. This paper proposes a new method for clustering and visualizing alternatives in the context of preference-approvals. Firstly, we present a new family of pseudometrics defined on a set of alternatives, evaluated via preferenceapprovals. The distances among alternatives are used as input in the Ranked k-medoids algorithm to find clusters. Finally, clusters are visualized in a two-dimensional space using non-metric multidimensional scaling. We show, through an application to real data, that our approach allows for reducing the complexity of the preference-approval space and facilitates its interpretation.