Improvement And Application Of Artificial Bee Colony Algorithm

As a new swarm intelligence optimization algorithm, artificial bee colony algorithm developed very rapidly in recent years. In 2005, the Turkish scholar Karaboga proposed artificial bee colony base on principle based on honey bees principle and applied it to multivariate function optimization problems. Because artificial bee colony algorithm has fewer parameters, simple and fast convergence, etc., it soon attracted the attention of many researchers. New applications based on artificial bee colony algorithm are constantly proposed, no longer confined to the function optimization. From the current research papers, artificial bee colony algorithm has been successfully applied to the traveling salesman problem, the deployment of wireless sensor network nodes, scheduling problems, parameter optimization and image segmentation, and many other fields, and the new application scenarios continues to be proposed. However, in the application process, the researchers also found that the shortcomings of artificial bee colony algorithm. In the face of complex optimization problems, artificial bee colony algorithm is easy to fall into local optimal and the convergence rate is also not satisfactory. Therefore, improving artificial bee colony algorithm has become a hot research direction. The main contents of this paper are as follows:First, for the shortcomings of the basic artificial bee colony algorithm, this paper proposes an improved artificial bee colony algorithm. There are two main improvements. First, improve the nectar updating formula. In basic artificial bee colony algorithm, when updating the nectar, honey bees search around the goal nectar randomly, the blind search leads to slow convergence. In this paper, based on mutation operator of differential evolution algorithm, we use neighbor nectar as a guide when searching new nectar, and introduce the random dislocation crossover strategy to absorb beneficial information on other dimensions. The new nectar updating formula is more purposeful, and it has stronger local search ability. Second, improve the nectar fitness formula. The fitness formula of basic artificial bee colony algorithm sometimes can not truly reflect the quality of honey, which seriously affects the accuracy of the optimization algorithm. In order to better evaluate the nectar and protect the diversity of population, this paper proposes a new sort-based fitness formula. In order to demonstrate the effectiveness of the improved algorithm, with the same parameter settings, we use the improved algorithm, the basic algorithm and other improved algorithms to optimize a representative set of standard test functions, compare and analyze the experimental results. Experiments show that the improved artificial bee colony algorithm converges faster and has more accurate results.Secondly, the artificial bee colony algorithm is applied to the graph vertex coloring problem. The basic artificial bee colony algorithm is mainly used for continuous function optimization, but not suitable for solving the constrained combination optimization problems. According to the specific needs of the graph vertex coloring problem, we redesign the target function and the nectar updating formula, propose an artificial bee colony algorithm for solving graph vertex coloring problem. We use integer arithmetic to code the feasible solution. A feasible solution represents a coloring scheme, it is expressed by a vector whose length is the total number of the vertexes. Each dimension of the vector represents the color of the vertex. The objective function value of feasible solution is smaller, the better the coloring scheme is. Therefore, to find the optimal coloring scheme is to find the feasible solution which makes the objective function obtain the minimum value. The nectar updating operations is to change the color of a vertex randomly in the premise of not increasing the total number of the colors. In the experimental stage, we use several standard cases to test the algorithm.