Tropical implementation of the Analytical Hierarchy Process decision method

Abstract

We apply methods and techniques of tropical optimization to develop a new theoretical and computational framework for the implementation of the Analytic Hierarchy Process in multi-criteria problems of rating alternatives from pairwise comparison data. In this framework, we first consider the minimax Chebyshev approximation of pairwise comparison matrices by consistent matrices in the logarithmic scale. Recasting this approximation problem as a problem of tropical pseudo-quadratic programming we then write out a closed-form solution to it. This solution might be either a unique score vector (up to a positive factor) or a set of different score vectors. To handle the problem when the solution is not unique, we develop tropical optimization techniques of maximizing and minimizing the Hilbert seminorm to find those vectors from the solution set that are the most and least differentiating between the alternatives with the highest and lowest scores, and thus are well representative of the entire solution set.

37 pages (added a new example and a discussion). arXiv admin note: text overlap with arXiv:1801.10524

References in corpus (3)

• Duality and separation theorems in idempotent semimodules
• Using tropical optimization techniques to evaluate alternatives via pairwise comparisons
• Algebraic solution of tropical optimization problems via matrix sparsification with application to scheduling

Cited by in corpus (6)

• Algebraic solution of tropical polynomial optimization problems
• Using pairwise comparisons to determine consumer preferences in hotel selection
• Quasi--Euclidean classification of Alcoved Convex Polyhedra
• Application of tropical optimization for solving multicriteria problems of pairwise comparisons using log-Chebyshev approximation
• Algebraic solution to constrained bi-criteria decision problem of rating alternatives through pairwise comparisons
• Algebraic solution to box-constrained bi-criteria problem of rating alternatives through pairwise comparisons