A comparative study on clustering-based group scenario summarization in AHP

Group scenario summarization is a useful approach in group decision making. In order to construct intuitive summaries of AHP evaluation weights, this paper adopts several k-means-type clustering methods to AHP results. In criterion selection level, AHP weights on several criteria are summarized into interval weights for representing the tendencies of group preferences in each cluster. In alternative selection level, similarities among criteria are evaluated by comparing cluster tendencies in criterion-wise selections with the goal of merging familiar criteria. Through several comparative experiments, applicability of several clustering method such as noise rejection and k-member clustering is discussed.