Performance evaluation of self organizing genetic algorithm for multi-objective optimization problems

Abstract

Self Organizing Genetic Algorithm (SOGA) uses a weighted-sum fitness assignment approach for solving multi-objective optimization problems. SOGA has been developed based on minimum genetic algorithm (GA) requirement that is easier to implement and customized to other multi-objective problems. This paper presents the performance of SOGA in terms of convergence, diversity, and consistency using various selected multi-objective benchmark problems with different pareto front features. The performance of SOGA is also compared with other well known evolutionary methods such as NSGA-II, PESA and PAES. The results show that SOGA provided a good convergence and high consistency in most cases of problems. For the case of diversity, SOGA performance is inferior as compared with others. However, SOGA is still able to obtain many optimal solutions, which are distributed on the true Pareto front.