Performance Analysis Of An A Priori Strategy To Elicitate And Incorporate Preferences In Multi-Objective Optimization Evolutionary Algorithms

The project portfolio selection is one of the most important strategic problems, both in the private sector and in the public sector. This can become a complex activity due to several factors, as occurs in many real-world optimization problems in which many criteria must be considered simultaneously. The preferences of a Decision Maker (DM) are a relevant element for decision-making activities, in general, and in portfolio selection, in particular; they vary between decision-makers and evolve over time. A strategy is required that assists the DM in the identification of the best compromise solution that satisfies their preferences. In order to incorporate DM's preferences, given in examples, the methodology Preferences Disaggregation Analysis (PDA) is introduced to obtain the parameters of a preference model from examples. This model is the basis of a classifier that allows to a multi-objective optimization evolutionary algorithm lead the search towards the DM's region of interest. In this paper is analyzed the performance of two multi-objective optimization algorithms of the state of the art when preferences are elicited indirectly through a PDA method. The experimental results showed the potential of the proposed method applied to small and medium scale instances.