Application And Research On The Bees Algorithm

A plenty of complex optimization problems that existed in the various sciences and real life is greatly meaningful for the development of sciences and practical application. However, the solution of this sort of problem that heavily relies on some high-performances algorithms of the optimization that is different from traditional ones, is Bees Algorithm that imitates feeding behaviors of the wasp swarm as one of the most successful evolutionary algorithm specially for complex optimization problems during the past decades.Bees Algorithm as one of the Swarm Algorithms is created by D.T.Pham and A.Ghanbarzadeh from the university of Cardiff in 2005. The principle of Bees Algorithm formed from feeding behavior is to require individuals with high fitness to conduct local searches in the closed domain and ones with lower fitness to finish global searches in the whole domain according to order of the fitness. This method increases the rate of convergence tremendously through exploring the most potential accessible solutions. After this successive processes, this algorithm would meet the requirement of evolution generations to find similar solution. However, many researches has claimed that some disadvantages of this algorithm exist such as low-efficiency rate of convergence, low-quality solution and so on although sometimes this algorithm is better than other else relatively.This theme is to improve the performance of the algorithm specially based on the disadvantages of Bees Algorithm and conduct enough exercises for classified problem in the revolution model and convergence problems. Also, this article would try to make enough contribution to this kind of algorithm research.The main work in this paper as follows:(1) First this paper introduces background and significant of the topic, reviews the current situation of Bees Algorithm. Then this paper introduces idea of Bees Algorithm and specific content of the main steps of the algorithm.(2)This paper introduces evolutionary algorithm for solving classification problems and principles. Grammar Evoluntion algorithm is lead into improved Bees Algorithm. A new improved algorithm called Grammatical Bees Algorithm(GBA)is proposed and uses to solve classification problems. By using several standard data sets for simulation experiment results of the study show that: Compared with the basicgene expression programming classification algorithm and improved GEP classification algorithm, GBA can get a better classification accuracy and the convergence rate increase greatly, which illustrates the feasibility and effectiveness of the GBA.(3) Based on the framework of Bees Algorithm introduces three new evolution operators,exchange operator and edge mutation operator as well as gradient global contraction operator,thus forming an improved Bees Algorithm which can be used to solve one hundred circles envelope issue of unequal circle packing problem likewise.Experimental results show that there is greater improvement to obtain the optimal solution of this problem by using this algorithm than other current evolutionary algorithms,which indicates the feasibility and effectiveness of the new algorithm.