A comparative study of the state-of-the-art algorithms on multi-objective problems using performance metrics

Abstract

Most real-world optimization problems have several objectives which can require to satisfy simultaneously. As the complexity of the problems increases, the problems become more challenging to solve for the algorithms. Over the last few years, many state-of-the-art multi-objective optimization algorithms (MOOAs) have been developed but, to the best knowledge of the authors, there has been no comprehensive comparative study in the literature together with these algorithms. Moreover, there are many parameters that affect the performances of algorithms, such as the population size, iteration number, and initial population. These values may also differ from one study to another, making comparisons more difficult to perform fairly. In this study, four MOOAs which were not previously used together for the comparisons of performance are applied to find optimal solutions of various mathematical optimization problems with the same initial conditions. The performances of the algorithms are evaluated by using two different performance metrics. In addition, Pareto fronts found by the algorithms are given comparatively. Experimental results give an overview for the performance of each algorithm on the different type of the problems by making fair comparison.