USING GENETIC ALGORITHM IN FOREST PLANNING CONSIDERING ITS SELECTION OPERATORS

• Autor(a) principal: Gomide, Lucas Rezende
• Data de Publicação: 2015
• Outros Autores: Arce, Julio Eduardo, Silva, Arinei Carlos Lindbeck da
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Cerne (Online)
• Texto Completo: https://cerne.ufla.br/site/index.php/CERNE/article/view/180
• topic: Metaheuristic; combinatory analysis; model type I

This study tested and analyzed four selection operators (Elitist, Tournament, Roulette wheels and Bi-classist) and defined the best one.  The forest planning problem test was based on the Johnson & Schermann (1977) type I model encompassing 52 eucalyptus stands, where 254 forest management prescriptions were created. The genetic algorithm (GA) was built in Visual Basic® Microsoft® and its sets of parameters were: initial population (300), crossover (10%), mutation (10%) and replacement (60%). The measuring variables were: minimum, median and maximum values; coefficient of variation for the fitness and the processing time.  It was also applied the nonparametric Kruskal-Wallis test with 5% of the probability to check the differences among selection operators of 30 samples. The results showed that the selection operators presented different efficiency and effectiveness according to Kruskal-Wallis test for 5% of probability. The decreasing sequence of efficiency was: Roulette wheels, Tournament, Elitist and Bi-classist. The lower percentage deviations matched from the exact solution were: 2.75% (Elitist), 2.15% (Tournament), 0.90% (Roulette wheels) and 2.40% (Bi-classist). The best selection operator tested was the one that follows the Roulette wheels.