Preference Robust Ordinal Priority Approach and its Satisficing Extension for Multi-Attribute Decision-Making with Incomplete Information
Ordinal Priority Approach (OPA) has recently been proposed to deal with multi-attribute decision-making (MADM) for determining the weights of experts, attributes, and alternatives under incomplete preference information. This study presents an equivalent reformulation of OPA related to rank order ce...
Gespeichert in:
1. Verfasser: | |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | Ordinal Priority Approach (OPA) has recently been proposed to deal with
multi-attribute decision-making (MADM) for determining the weights of experts,
attributes, and alternatives under incomplete preference information. This
study presents an equivalent reformulation of OPA related to rank order
centroid weights, further providing its closed-form solution and
decomposability. Building on these properties, we propose a Preference Robust
Ordinal Priority Approach (OPA-PR) utilizing a two-stage optimization framework
to generalize the utility structure and counter the ambiguity within the
ranking parameters and utility preferences. In the first stage, OPA-PR elicits
the worst-case utility functions across all experts and attributes from utility
preference ambiguity sets characterized by monotonicity, normalization,
concavity, Lipschitz continuity, and moment-type preference elicitation. In the
second stage, OPA-PR optimizes decision weights based on the elicited utility
functions, considering the worst-case ranking parameters indicated by support
functions. We suggest a piecewise linear approximation for the utility
preference ambiguity sets to derive the tractable reformulation of OPA-PR,
verified by the error bounds for optimal outcomes in both stages. Critical
properties of OPA-PR are provided, including closed-form solution, invariance
of optimal weight disparity, and risk preference independence. Moreover,
considering the potential ranking parameter misspecification, we develop a
robust satisficing extension of OPA-PR based on its closed-form solution,
OPA-PRS, with a novel decision criterion-fragility measure of weight disparity.
The effectiveness of OPA-PR and OPA-PRS is validated through a numerical
experiment of the emergency supplier selection problem. |
---|---|
DOI: | 10.48550/arxiv.2412.12690 |