Sistemas Fuzzy genéticos baseados em regras: um operador de seleção com foco na diversidade das soluções

Abstract

The automatic generation and optimization of Fuzzy Systems is an important area because the manually generation and optimization is very hard and may require one or more specialists. Even with the help of specialists the work may be impossible according to the variables involved and the complexity of the problem. One of many possibilities to help solve this problem is the use of Inteligent Systems. The Multiobjective Genetic Fuzzy Systems are one example of Intelligent Systems that use Multiobjetive Genetic Algorithm. The propose of the Multiobjetives Genetics Algorithms is generate solutions that satisfies more than one objective at once. The selection of the individuals that will generate the next offspring depends on the Multiobjetive Genetic Algorithm used, each one of them uses a different method. The propose of this work is to implement an alternative selection method for the Multiobjective Genetic Algorithm (2+2) M-PAES. This method aimed to improve the dispersion of the solutions but it has improved the accuracy of the results and has keeped the interpretability of the Fuzzy System.