Influência dos parâmetros da metaheurística algoritmo genético em um problema de planejamento florestal

Abstract

This work aimed to evaluate the behavior of different parameters of the Genetic Algorithm metaheuristic to solve problems of forest planning. The study was developed considering a planning horizon of 16 years, analyzing as parameters of the metaheuristic the use of elitism, the type of crossing (1 cut-off point and uniform), the type of parent selection for crossing (roulette and tournament ) and mutation type (random gene choice and gene to gene), totaling 16 combinations. In addition, the initial population size (20, 50 and 80 individuals) and the stopping criterion (100, 300 and 500 generations) were evaluated, totaling 9 combinations. Each of these combinations of genetic algorithm parameters was considered as a different treatment and processed with 30 replicates. The problem was to find a scenario of forest production planning that would return the maximum Net Present Value subject to the restrictions of age of cut (between 5 and 7 years), minimum and maximum annual demands (140,000 m³ and 160,000 m³, respectively) and integrity. Processing was done using MeP software. To compare the results, the non-parametric Kruskal-Wallis test was used. The results show that there is a statistically significant difference, at 5% probability, between the parameter combinations. We conclude that the quality of the solutions generated by the metaheuristic genetic algorithm is dependent on the selection of a good configuration of its parameters. The best solution was found using the following configuration: elitism, 1-point crossing, tournament-type selection, gene-to-gene mutation, initial population with 50 individuals, and 500-generation stop criterion.