A Dichotomous Approach to Reduce Rank Reversal Occurrences in PROMETHEE II Rankings.

Multicriteria methods based on pairwise comparisons suffer from rank reversal occurrences when the set of alternatives is modified. In this paper, we consider an alternative PROMETHEE-based method to compute pre-orders. This should lead to face less rank reversal (RR) occurrences than in the traditional Promethee II ranking. It is based on hierarchical clustering using a dichotomous process. The property on which is based the dichotomous separation is explained. The two methods are compared. At first, we show that the rankings obtained are similar. Then, a comparison between the frequency of rank reversal occurrences in both methods is conducted.