Comparative Analysis of Multi-objective Metaheuristic Algorithms by Means of Performance Metrics to Continuous Problems

Abstract

Multi-objective problems (MOP) are an important type of real-world problem that involves more than one objective that must be optimized simultaneously. There exists a wide variety of metaheuristic algorithms aimed to work on that type of problems, however, the selection of which algorithm should be used to a given MOP depends on the expertise of the researcher. This decision is not straightforward and usually means to use extra computational resources to try different multi-objective metaheuristics even before to try to solve the interested domain. In the state of the art, there exists several metrics to compare and contrasts the performance of two or more given multi-objective algorithms. In this work, we use these metrics to compare a set of well-known multi-objective metaheuristics over the continuous CEC 2009 benchmark with the objective to give the interested researcher useful information to properly select a multi-objective algorithm.