Artificial Bee Colony Algorithm Guided By Elite Knowledge

With the progress of science and technology,evolutionary algorithm plays an increasingly important role in engineering practice.Artificial bee colony algorithm is a potential evolutionary algorithm,which has the characteristics of simple structure,less parameters and easy implementation.However,the basic artificial bee colony algorithm has weak exploitation ability,which leads to its slow convergence speed and the accuracy of the solution insufficiently in solving some complex problems.Elite knowledge is the knowledge of the problem that the elite individuals get through self-learning in the process of evolution.If individuals can make use of elite knowledge on the basis of self-learning,it will help to enhance the exploitation ability of the algorithm.Based on the learning mode of elite knowledge guidance,major contributions of this paper are summarized as follows.(1)Elite center knowledge guided artificial bee colony algorithm(ECABC)is proposed.The basic artificial bee colony algorithm exhibits slow convergence rate due to its strong exploration ability but insufficient exploitation ability when solving some complex optimization problems.To tackle this problem,ECABC uses the elite knowledge to determine the combination weight to construct the elite center individual,and designs the search strategies guided by the elite center individual to guide the search direction.In the test experiment,several variants of ECABC are designed to verify the effectiveness of the new search strategies.Moreover,ECABC is compared with eight existing artificial bee colony algorithms,six differential evolution algorithms and five particle swarm optimization algorithms.Experimental results demonstrate that ECABC can enhance the exploitation ability of search strategies,and the convergence rate and the accuracy of algorithm can be improved.(2)Artificial bee colony algorithm with dual elite knowledge guidance(DGABC)is proposed.The search strategy of the basic artificial bee colony algorithm lacks guidance results in weak local search ability,so DGABC proposes a guiding individual construction mechanism,in which the core individuals in the population utilize the elite knowledge to construct the core dominant individuals,and the members in the current neighborhood utilize the elite knowledge to construct the neighborhood leading individuals.Based on the guidance of the core dominant individual and the neighborhood leading individuals,a search strategy with dual guidance is designed to improve the local search ability of the algorithm.Thirteen classic benchmark functions and fifteen CEC2015 shifted and rotated problems were used to test the performance of DGABC,and the experiments show that DGABC is a new competitive algorithm.(3)Compare ECABC and DGABC in numerical experiments.These two new algorithms based on the learning mode of elite knowledge guidance are compared in terms of convergence accuracy and operation efficiency.Experimental results demonstrate that ECABC and DGABC both have good performance in the test function,and DGABC has more advantages in solving the unimodal problems or multimodal problems.ECABC and DGABC have their own advantages in the complex problems with shifting and rotation.ECABC runs faster on most problems.