Tuning genetic algorithm parameters using design of experiments

Tuning evolutionary algorithms is a persistent challenge in the field of evolutionary computing. The efficiency of an evolutionary algorithm relates to the coding of the algorithm, the design of the evolutionary operators and the parameter settings for evolution. In this paper, we explore the effect of tuning the operators and parameters of a genetic algorithm for solving the Traveling Salesman Problem using Design of Experiments theory. Small scale problems are solved with specific settings of parameters including population size, crossover rate, mutation rate and the extent of elitism. Good values of the parameters suggested by the experiments are used to solve large scale problems. Computational tests show that the parameters selected by this process result in improved performance both in the quality of results obtained and the convergence rate when compared with untuned parameter settings