Improved Artificial Bee Colony Algorithm And Related Application Research

With the development of science and technology,optimization problems have become more and more important.Optimization methods have been applied in various fields,such as function optimization,image processing,robot control and industrial production.However,optimization problems in these fields usually demonstrate the characteristics of multimodal,multidimensional,non-differentiable and constrained.Traditional mathematical optimization methods can not solve such problems well.On the contract,in recent years,by achieving an acceptable result in reasonable time on the problems,meta-heuristic optimization algorithms attract more attentions.Among the algorithms,artificial bee colony algorithm(ABC)shows a relative strong global search ability.The paper focuses on ABC and analyzes its advantages and disadvantages.Specifically,the paper improves the algorithm from the perspectives of dimension update strategy,initialization strategy,relationship between variables,iterative formulas,and optimal global and local search strategy.In addition,the paper applies improved algorithms many fields such as industrial production,communication network management,robot path planning,and feature selection.The main contributions as follows:(1)In view of the slow search speed and low accuracy of the traditional ABC algorithm,but it is not easy to fall into the local optimum when dealing with complex optimization problems,the paper both improves the search strategy of employed bee and onlooker bee to balance the exploration and exploitaion ability of the algorithm.Firstly,this paper analyzes the influence of updating dimension on the algorithm,proposes a reasonable linear updating dimension mechanism and applies it in the employed bee phase,which ensures the accuracy of early search and high speed of searching for better food sources.In addition,the high quality guidance strategy and the worst dimension updating strategy are integrated into the onlooker bee iteration formula to improve the algorithm accuracy.Finally,the algorithm is applied to the loudspeaker model design problem of electromagnetic device optimization and controller placement problem in software defined networks.Experimental results of the basic functions and the real world problems demonstrate the effectiveness of the algorithm.(2)ABC shows a strong global search ability but slow convergence speed.The disadvantage is more obvious in dealing with dimension-dependent problems.In this paper,an improved artificial bee colony algorithm is designed to enhance the effect of dimension-dependent parts.The paper analyzes the properties of search formulas of both the differential evolutionary algorithm and the artificial bee colony algorithm,and designs effective update strategies for the dimension-dependent and dimension-independent problems respectively,and proposes a co-evolution method to merge the two strategies.In the paper,fourteen classical dimension-independent benchmark functions and dimension-dependent CEC2015 functions are employed to verify the performance of the improved algorithm.Finally,the effectiveness of the algorithm is verified by using the robot path planning problem.(3)Focus on the discrete optimization problem,a multi-filter hybrid feature selection algorithm model based on discrete artificial bee colony is proposed to identify patients with Parkinson's disease in the paper.The model uses three different filter methods to initialize the population,calculates the fitness based on the classification accuracy and the number of selected features,and automatically selects the optimal classifier.The experimental results show that MFABC is an effective feature selection algorithm,which not only reduces the size of feature subsets,but also improves the classification performance of feature subsets.