Uma nova meta-heurística adaptativa baseada em vetor de avaliações para otimização de portfólios de investimentos

Abstract

This paper presents a metaheuristic based on evaluation vectors named AVEMH to solve multiobjective problems. This meta-heuristic consists of evolving two populations independently and exchanging information between them so that the first population evolves according to the best individual in the second population and vice versa. The choice of algorithms to be executed in each generation is stochastic between three evolutionary algorithms in the literature: PSO, DE and ABC. The results of this metaheuristic are compared using benchmark functions as well as applied to the real world problem known as Portfolio Selection. To evaluate the quality of the results obtained, we use its well-known metric in the analysis of evolutionary multiobjective algorithms called hyper volume. Tests have shown that the new metaheuristic can find the best hyper volumes in both the benchmark ZDT functions and the optimization of investment portfolios. Analysis of variance (ANOVA) and Tukey tests were applied to demonstrate that the new meta-heuristic can find better hyper volumes in both Zdt benchmark functions and investment portfolio optimization.