Bat Algorithm(BA)is a kind of swarm intelligent optimization algorithm which is simulated by bats' hunting behavior,its purpose is to get the global optimization,BA is superior to genetic algorithm and particle swarm optimization in performance,with the characteristics of easy to operate,less parameters and strong robustness,BA is help for solve different kinds of combinatorial optimization or continuous optimization problems,and it can be used to many fields,such as image processing,the allocation of resources,data mining,information processing,financial and so on.However,for solving some of the more complicated problems,the algorithm's defects are still exist,such as lack of population diversity,low precision,slow convergence speed and easy to fall into local optimization.This paper analyzes bat algorithm's steps of optimization and the bionic principle,compares it with different kinds of intelligent optimization algorithm,summaries the advantages and disadvantages of BA in solving different optimization problems,and proposes an improved bat algorithm based on Opposition Learning strategies(OLBA).In order to solve the problem of low population diversity,using uniform design theory to initialize the population,which can make all individuals distributed relatively in the search space,so as to construct a better population diversity.Aiming at the problem of low precision and slow convergence speed,using Opposition Learning strategies,by which elite bats are introduced to generate their opposite population,and then select fitter individuals from current population and the opposite population,thus optimizing the individual's position in bat's population.Adding adaptive formula to BA's updating formula to achieve the goal of adjusting the moving step dynamically,so that it can target the optimal target better;for the problem that plunged into local optimization easily,set the threshold in judging population diversity,when the threshold is not satisfied,initialize some individuals who have poor fitness values in the search space in order to change their positions,which can improve the diversity of population and avoid the premature convergence of the algorithm effectively.This article uses six kinds of unimodal function and four kinds of multimodal functions to simulate BA and OLBA respectively.The experimental results show that OLBA is superior than BA in the optimization process,which can accelerate convergence,get finer solution when BA trapped in local optimal solution,and improve the accuracy of solution to a largeextent.In this paper,six unimodal functions and four multimodal functions are used to simulate BA,Bat algorithm with the characteristics of Lévy flight(LBA)and OLBA respectively.The comparative and analysis of the experimental results can be concluded that OLBA can accelerate the convergence speed,get more detailed solution in the process of optimization,and get the golbal optimal solution stably.In summary,OLBA's improvement strategy is effective and feasible.The main work of this paper is to improve bat algorithm according to its shortcomings.The keys to affect the algorithm performance include setting up reasonable initial population,making rules to judge population diversity,operating the elite individual,updating formula,these factors will be studied further in subsequent work,such as the selection of mobile factors in the individual update formula,the threshold of population diversity and so on,and I will detailed analyze other groups of intelligent optimization algorithm to get breakthrough points,tap the potential of the bat algorithm,and strive to further improve the performance of BA,and apply it to reality better. |