The Improving Study On Cuckoo Search Algorithmand Bat Optimizationalgorithm

The cuckoo search algorithm and bat optimization algorithm are both swarmIntelligence algorithm, which are proposed by Yang who come form CambridgeUniversity. The cuckoo search algorithm derives from the obligate brood parasiticbehaviour. The algorithm has been successfully applied to the neural networktraining, face recognition, engineering design optimization and mult-objectiveoptimization. At the same time, the bat optimization algorithm derives from batpredatory behavior and echolocation. The algorithm has already obtained certainsuccess application in fuzzy clustering, neural network training, numericaloptimization problem, and so on. The two algorithms has become a hot researchscholars at home and abroad. However, the cuckoo search algorithm has somedeficiencies, such as local search is not strong, low precision, and so on. The batoptimization algorithm is a new swarm intelligence optimization algorithm, whichhas the problems of the premature convergence and the theoretical foundation isnot perfect.This paper makes some researches of improving the performance of the twoalgorithms, the mainly results include the following aspects.(1)According to the problems of cuckoo search algorithm, such as localsearch is not strong, converge slowly in the later period and low precision, a newcuckoo search algorithm based on elimination mechanism(EMCS) is proposed.The experiment indicated that EMCS has good performance in converge rate andprecision.(2)To avoid premature convergence and improve global convergencecapability, a modified bat algorithm by using maneuver flight (MFBA) is proposed.The experiment show that MFBA can avoid the premature convergence and has astrong global searching capability.(3)In order to improve the theoretical foundation and the capability ofhigh-dimensional optimization, this paper improve the update formula of the speed and position. An improved bat algorithm with memory characteristic (MCBA) isproposed. Experiment results show that improved algorithm has clear superiority.