GENETIC ALGORITHM PARAMETER CONTROL FOR ACHIEVING BETTER OPTIMIZATION PERFORMANCE

Abstract

This research is directed towards controlling genetic algorithm operator parameters. Simulations have been done in MatLab on examples taken from literature for genetic algorithm testing. Based on a large number of simulations with different parameter values, algorithm operator values are attained experimentally. An analysis of results has been completed in Statistica, as well as the creation ofž nonlinear equations for the correlation between operators and results. By optimizing the derived equations it is possible to determine general parameter values of operators which will have beneficial optimization performances, in terms of convergence. One equation which gives the best optimization values is favored. Attained values are again tested on new examples which define achieved performance and benefits of this approach. These results lead to a simplified use of the genetic algorithm for practical optimization with satisfactory results. This approach has a practical engineering optimization use perspective.