Visualization-aided Multi-criterion Decision-making Using Reference Direction Based Pareto Race

Abstract

The goal of a multi-criteria decision-making (MCDM) approach is to select one or a few preferred solutions in an iterative manner from a set of Pareto-optimal solutions obtained by a generative or simultaneous evolutionary multi-and many-objective optimization (EMO and EMaO) algorithm. In each iteration, the decision-maker (DM) formulates a suitable scalarized optimization problem using preference information that guides the DM to arrive at the most desirable solution set. Visualization of trade-offs among multiple objectives and their interactions with constraints can provide crucial decision-making information. In this paper, we propose a visualization-assisted MCDM approach that utilizes interpretable Self-Organizing Maps (iSOM) on a well-known MCDM technique known as Pareto Race. The proposed method, applied to one test and two real-world problems involving three to five objectives, demonstrates the usefulness of the iSOM-visualization method in implementing Pareto Race decision-making approach. The study opens up further avenues for integrating iSOM visualization approach with other MCDM techniques.