A New Approach Method of Crossover Process Based On Genetic Algorithm Using High Dimensional Benchmark Functions

Abstract

The design of the improved genetic algorithm (GA+) is based on a meta-heuristic search for optimization problems. In this paper, the crossover process in the original genetic algorithm is improved. The improvement of the crossover process is renewed by applying two conditions. One of them is keeping the last genes (constant) for each population; the second one is about rotating genes according to the defined range of points between each two selected populations. The improved genetic algorithm (GA+) has the possibility of accelerating local convergence. Therefore, it gets a chance to search for better values globally using these conditions. All processes in the improved genetic algorithm have been represented in this paper. The performance of the proposed algorithm is evaluated using 7 benchmark functions (test functions) on different dimensions. Ackley function, Rastrigin function and Holzman function are multi-modal minimization functions; Schwefel 2.22 function, Sphere function, Sum Squares function and Rosenbrock function are uni-modal minimization functions. These functions are evaluated by considering cases that are minimized by having a set of dimensions as 30, 60, and 90. Additionally, the performance of the GA+ is compared with the performance of comparative optimization algorithms (meta-heuristics). The comparative results have shown the performance of the GA+ that performs much better than others for optimization functions.