Impact of Parent Selection Operator on the FDEA Algorithm

• João Pedro A. de Oliveira UFAC
• Olacir R. Castro Jr. UFAC

Resumo

Many-Objective Optimization Problems (MaOPs) present a significant challenge due to the increased complexity associated with optimizing more than three conflicting objectives. Traditional Pareto-based algorithms often struggle to scale effectively with the number of objectives, leading researchers to explore alternative approaches. This study investigates the impact of parent selection strategies on the performance of the Fuzzy Decomposition-Based Multi/Many-Objective Evolutionary Algorithm (FDEA). The FDEA uses a fuzzy decomposition approach to divide a MaOP into subproblems, each solved individually to improve the adaptability and precision of solutions. The study compares the FDEA with three selection methods: binary tournament based on Pareto dominance, binary tournament using crowding distance, and a combination of Pareto dominance and crowding distance. Empirical evaluations were conducted using the DTLZ and WFG benchmark problem sets, with 2, 3, 5, 8, 10, and 15 objectives. The results indicate that incorporating quality indicators into the selection process improves the efficiency and effectiveness of the FDEA, especially in high-dimensional optimization scenarios. However, no variant was superior in all contexts. The study highlights the potential for methodological advances in Indicator-Based Mating Selection, emphasizing the need for robust and scalable algorithms capable of handling the complexities of MaOPs.