Study on a novel crowding niche genetic algorithm

Abstract

This paper proposes a new crowding niche genetic algorithm to make up the shortages of bad stability, poor local search ability, and inferior universality in conventional crowding niche genetic algorithms. The new algorithm develops a new crowding strategy based on the most similar individuals to maintain the population diversity, designs an improved mutation probability adaptive adjustment method in accordance with the change law of sigmoid function curve to accelerate the convergence speed, and introduces the gradient operator into computation process to enhance the local search capability. Four typical complex functions are selected as test functions and two conventional algorithms are applied as contrast algorithms to assess the performance of algorithm. Test experiments and comparative analysis show that the proposed algorithm has an outstanding performance for maintaining population diversity; it is very effective and universal for solving complex problems. The new algorithm generally outperforms conventional crowding niche genetic algorithms.