Modified Artificial Bee Colony Algorithm For Search Ability Enhancement

Artificial bee colony algorithm is a kind of swarm intelligence optimizationalgorithm, the simulation of the honey bees. The algorithm according to the differentdivision of labor in exchange for nectar source of bees information, search for theoptimal solution of optimization problem. Because of its less set parameters, simplecalculation, and the advantages of easy to implement, has been successfully applied infilter design, network optimization and production scheduling, etc. But because of theswarm algorithm uses the roulette wheel selection mechanism to search update of honey,the bees diversity decreased, algorithm premature stagnation, and is easily plunged intolocal optimal problems.In view of the shortcomings of algorithm, the theory optimum solution ofprobability to be increased, the paper research on original colony algorithm and it isimproved, gaining a modified artificial bee colony algorithm for search abilityenhancement. The improved algorithm will make the number of onlooker bees isdoubled and it is divided into two groups, one of the groups uses the roulette wheelselection strategy update nectar source location, keep the direction of the evolution ofthe original, along the direction of high concentration of nectar sources search updates;Another group uses reverse roulette selection mechanism, maintain the swarm diversity,to reduce the probability of local optimum. Then, select all the nectar sources which thetimes of counter without updating is greater than the threshold value, and these nectarsources corresponding employed bees all become scout bees, and these nectar sourcesare updated. By the adaptive dynamic adjusting the number of scout bees, to improvethe convergence precision of the algorithm. After each iteration, the swarm algorithm tosolve the center position of nectar sources, a new solution is produced in theneighborhood, formed by the center position and every nectar source, and the fitnessvalues are compared to choose the optimum. Through six standard test functions ofexperimental simulation, the results show that the algorithm has higher convergenceaccuracy, and increase the opportunity to jump out of local optimal algorithm,strengthen its optimization ability, has better optimization performance.Finally, the paper summarized the research work and achievements of overview,andgive clear instructions to the next research direction.