Improved Artificial Bee Colony Algorithm Based On Multiple Swarms

In recent years,swarm intelligence optimization algorithm simulates biological group living,reproduction and other behaviors,replaces global search with local random search,and replaces definite solution with approximate solution from local to global,from individual to group.Compared with other optimization algorithms,swarm intelligence optimization algorithm has a good performance.Among them,artificial bee colony algorithm is proposed by simulating the foraging process of bees.Due to its advantages such as simple structural framework,better structure and fewer parameters,it has great advantages compared with other swarm intelligence optimization algorithms and has been favored by a large number of scholars.In this paper,the artificial bee colony algorithm has the problems of slow convergence speed and easy to fall into local optimum.Through in-depth research and analysis,it is found that the main cause of the problem is the mismatch between the exploration ability and development ability of the population.A large number of studies and experiments show that population diversity is the guarantee of algorithm exploration ability,so this paper focuses on population diversity to improve the development ability of the algorithm,and at the same time to ensure the diversity of the population to improve the algorithm exploration ability.Based on the above considerations,this paper proposes a multi-population division,which guarantees the diversity of the population through the support of multiple sub-populations.On the basis of multi-population,two improved artificial bee colony algorithms are proposed.The main work is as follows:1.Divide the population into multiple sub-populations of equal size,and divide them into excellent,good and poor grades according to their respective performance.In the employed bee phase,three different search strategies were proposed to improve the exploration ability of excellent subpopulations,the exploration ability of poor subpopulations,and the balance between exploration and exploitation of good subpopulations.In the onlooker bee phase,the global optimal solution and the elite solution are introduced,two search equations are proposed,and the current optimal strategy is selected according to the combination of current search experience and past experience.Finally,in order to improve the power of the algorithm,two probability models are established to select subpopulations and individuals respectively.In individual selection,the COS curve is proposed to ensure the selection of the optimal individual and improve the selection probability of the worst individual,thus providing the algorithm with the ability to jump out of the local optimal.2.Based on the algorithm of the first one,the size and quantity of subpopulation are considered in this paper.The fixed and equal size of each subpopulation may lead to the fact that some individuals are not suitable for the current subpopulation,and the internal differentiation of the subpopulation is too large,so that some individuals can not play their role.Therefore,in the second algorithm,the subpopulation size is not equal,and the most suitable subpopulation size and number are selected by the algorithm adaptively.In addition,as the number of subpopulations changes with the number of individuals,the exploration and development ability within the size range also presents great differences,so the corresponding probability selection model is also adjusted accordingly.The results of the above two algorithms in the two test sets all verify their effectiveness in improving the artificial bee colony algorithm and ensure the balance between algorithm exploration and development.