Performance Evaluation of a Hybrid NSGA-III for Multi and Many-Objective Optimization in Real-World Problems

Several real-world problems can be modeled as optimization problems in which multiple conflicting objectives must be optimized simultaneously. Evolutionary algorithms (EA) are able to identify a set of non-dominated solutions (Pareto front) and are commonly used to solve these problems. Hybrid EAs that combine various optimization techniques can leverage the strengths of each method involved, enhancing their overall effectiveness. This study evaluates the performance of the well-known NSGA-III hybridized with Differential Evolution, Sine Cosine, and Arithmetic algorithms on optimization problems based on real-world applications with three, four and five objective functions, extracted from the CEC 2021 competition. Performance comparisons between the hybrid and original versions of NSGA-III were conducted using the IGD+ and HyperVolume indicators. Statistical analysis via the Wilcoxon test revealed significant improvements in NSGA-III performance when the hybrid version is considered.