The improvement of data in pairwise comparison matrices

Abstract This article analyzes the question of reliability of entries collected in the pairwise comparison matrices used for prioritization of alternatives. The inappropriate input data may result in an invalid output hierarchy. Basing on the Geometric Mean Method introduced by Crawford and Williams and the geometric consistency index defined by Aguaron and Moreno-Jimenez we introduce an algorithm which allows to search for the least reliable pairwise comparison matrix elements and improve them by means of their estimations.