NSGA-DO: Non-Dominated Sorting Genetic Algorithm Distance Oriented

In this work, a multi-objective genetic algorithm named Non-dominated Sorting Genetic Algorithm Distance Oriented (NSGA-DO) is proposed. It has been designed as a modification of the well known NSGA-II. The proposed algorithm is able to find non-dominated solutions that balance the Pareto front with respect to optimization of the objectives. The main characteristic of NSGA-DO is the distance oriented selection of solutions. At each iteration, the non-dominated solutions are used to find an approximation to the Pareto front. The algorithm uses the locations of the solutions in the approximated frontier to find the best distribution of solutions, which will guide the selection operations. In order to validate the proposal, NSGA-DO was applied in the context of Multi-Objective Evolutionary Fuzzy Systems (MOEFS), to the generation of fuzzy knowledge bases for classification. The study focus on the evaluation of the distribution of non-dominated solution as well as on the accuracy-interpretability trade-off. Experiments show the superiority of NSGA-DO when compared to NSGA-II in all three issues analyzed: dispersion of non-dominated solutions, accuracy and interpretability of the generated systems.