Improvement And Application Of Artificial Bee Colony Algorithm

Artificial bee colony algorithm is a novel meta-heuristic search algorithm. The principle of implementation simulates the intelligent foraging behavior of honeybee swarm to solve the practical problems. Because the artificial bee colony is simple and easy to understand, easy to implement and robust, and has less parameters, since it has been proposed by Turkish scholar Karaboga in2005, it has been successfully applied to solving constrained numerical optimization problem, multi-objective optimization problem, artificial neural network, the detection and prediction of protein, dynamic path selection, reliability redundancy allocation and so on. And all of them have achieved excellent results. However, as a novel algorithm, its model is not mature, and it is still in the initial stage in the applications of some complicated practical problems. Therefore, it has important research value and practical significance to improve the theory research and explore the applications of complicated practical problems. Through careful research will find that there are many deficiencies in artificial bee colony, such as slow convergence speed and easily falling into local optimal solution. These deficiencies of the algorithm make people unsatisfied when solving some problems. So many scholars committed to improve the algorithm so hard that the algorithm can be used to solve practical problems well and be used in a wider range of application.In this paper, by taking examples from the differential evolution mutation operator and in view of the shortcomings of artificial bee colony algorithm, we proposed several improved algorithms and applied one of the outstanding performance improvement to solving nonlinear equations. The main work is as follows.Firstly, the paper describes the origin, the biological model and the basic idea of artificial bee colony algorithm, and a detailed analysis of the basic steps, the time complexity and the characteristics of the algorithm. Then aiming at the slow convergence speed problem, we introduce the information of the current global optimal solution into the improved algorithm, so that we can effectively guide the search for fast convergence to the global optimum. In the simulation experiment, we can see the convergence speed of the improved algorithm is faster than the standard artificial bee colony algorithm in solving unimodal problems. Secondly, aiming at the easily falling into local optimum problem, we introduce one or two perturbation vector into the search way of the improved algorithm. This can maintain the diversity of the population and prevent the algorithm falling into a focal optimum. The experiment results show that the convergence precision of the improved algorithm is higher than other intelligent algorithms in solving multimodal problems, which is the presentation of jump ing out of the local optimum.Finally, we adjust the way to calculate the fitness of solutions in the original artificial bee colony, which is very complicated. We choose the direct use of the value of a function as the fitness of a solution for both simple and clear. And we choose one algorithm which performs better than others in the above improved algorithms, and apply it to solve nonlinear equations. The final experiment tests on a set of basic functions and a set of nonlinear equations. We make a longitudinal comparison with other swarm intelligence algorithm and a horizontal comparison with other improved artificial bee colony. The experimental results show that the improved algorithm is more suitable than the other algorithms in solving nonlinear equations.