Least Absolute Deviation Utility for Trapezoidal Fuzzy Preference Relations

Abstract

Preference relations (PRs) are widely used to model expert judgments because they allow for eliciting the decision-makers' opinions from pairwise comparisons. Traditionally, PRs have been elicited using real numbers. However, in real-world decision-makers usually feel more comfortable using linguistic expressions closer to natural language. In this context, our purpose is to extend the classical idea of PR into the environment of Trapezoidal Fuzzy Numbers (TrFNs) by addressing several drawbacks in current research. Existing fuzzy extensions for Fuzzy Preference Relations (FPRs) and Multiplicative Preference Relations (MPRs) assume that the notion of neutrality must be modeled by a crisp real number, which fails to capture the subjective and diverse ways in which decision-makers may perceive indifference. Moreover, current research lacks a theoretical framework that unifies both FPRs and MPRs for fuzzy numbers and simultaneously generalizes the involved classical notions (e.g., reciprocity, consistency) from the perspective of fuzzy arithmetic. Therefore, we introduce the notion of neutral TrFN, a fuzzy number that can model indifference on a certain fuzzy scale. This provides a flexible representation of neutrality, leading to a more realistic extension of PRs into the fuzzy environment. Building upon this, we propose the Trapezoidal FPR (TrFPR), analyze its properties and how reciprocity and consistency are extended using fuzzy arithmetic. Furthermore, we introduce the Least Absolute Deviation (LAD) utility vector associated with a consistent TrFPR and develop an optimization model to derive it from an inconsistent TrFPR to compute priority values. These ideas are also extended to MPRs in the fuzzy domain. Finally, the applicability of the proposed LAD utility method is demonstrated in evaluating land development projects, where comparison with fuzzy AHP illustrates its superiority.