Extension of the PROMETHEE Method to the Multicriteria Dual Clustering Problem

Multiple criteria decision making and clustering are topics that have been developed separately for decades. More recently researchers have investigated how to apply clustering techniques to multiple criteria decision aid. As in unsupervised classification, the goal is to obtain homogenous clusters that are well-separated. The distinctive feature comes from the fact that objects are compared based on preference relations (which are most of the time not symmetric unlike a traditional distance measure). In this contribution, we address the opposite problem. We want to find a partition of objects evaluated on multiple criteria such that groups obtained exhibit a high intra-group heterogeneity and good inter-group homogeneity. We present an extension of the PROMETHEE method to address this issue. This is illustrated on the problem of the creation of groups of students where fairness between the obtained clusters must be ensured.