Cuckoo search (CS) algorithm is a meta-heuristic optimization algorithm based on Lvy flights.We propose an improved cuckoo search algorithm to enhance the accuracy

avoid the local optima and accelerate the convergence speed.The proposed algorithm has three main characteristics.Firstly

pattern search method enhances the exploitation ability of the basic CS algorithm.The proposed algorithm combines the random exploration of the CS algorithm and the exploitation capacity of pattern search method.Secondly

the optimal solution is obtained by self-adaptive competition mechanism.Hence

the proposed algorithm has a trade-off between searching speed and the diversity of solution.Finally

we realize the cooperation of the optimal solution to share

and strengthen the advantage of experience learning in the use of the optimal solution set search mechanism.The experimental results conducted on 52 benchmark functions show that the proposed algorithm is promising in terms of accuracy

success rate and robustness.And it is also suitable for multimodal and high-dimensional numerical optimization problems.Therefore

in terms of the global search ability and solution accuracy

our algorithm performs better than other modified CS algorithms

such as ICS (Improved Cuckoo Search algorithm)

CSPSO (Cuckoo Search algorithm and Particle Swarm Optimization)

OLCS (Orthogonal Learning Cuckoo Search algorithm)

etc.