A Bayesian Framework for Attribute Weight Elicitation in Multiattribute Decision Analysis

Abstract

Multiattribute decision analysis requires evaluating alternatives based on multiple, often conflicting attributes, where weight elicitation plays a crucial role. Existing weight elicitation methods rely on distinct assumptions and lack a unified framework for incorporating diverse preference structures and handling uncertainty. This paper introduces a Bayesian framework for weight elicitation in multiattribute decision-making based on a probabilistic approach. The proposed framework provides statistically rigorous solutions for handling some key challenges in weight elicitation, such as attribute correlation and weight aggregation in group decision-making problems. Additionally, these models accommodate different forms of uncertainty in decision-makers’ preferences, including normal and triangular distributions, as well as interval preferences. Through experimentation on various numerical examples, the proposed framework is validated, demonstrating its effectiveness and highlighting its distinguishing features in comparison to alternative methods.

Keywords: Multiattribute decision-making, Attribute weight, Bayesian hierarchical model, Probabilistic ranking