An Enhanced Cuckoo Search Algorithm for Solving Optimization Problems

Abstract

In this paper, we present an enhanced cuckoo search (ECS) algorithm based on Gaussian diffusion random walks and greedy selection approach. Despite wide applications of cuckoo search (CS) algorithm, it suffers from low convergence speed and lacks balance between local and global search. To overcome these limitations, ECS algorithm is proposed. It employs Gaussian diffusion random walks instead of Levy flights random walks to enhance the local search. In addition, the greedy selection approach is employed to ensure that ECS reaches the optimum solution. Twenty one IEEE benchmark functions are used to evaluate the performance of the proposed ECS algorithm against CS and a recent adaptive cuckoo search (ACS) algorithms. The ECS shows excellent performance in reaching optimum values with a high convergence speed compared to CS and ACS algorithms. In addition, with several experiments, the effects of population size and the abandon probability $\pmb{P}_{\pmb{a}}$ are carried out for the three algorithms and experiments shown that ECS is more robust than CS and ACS algorithms. Furthermore, superior performance of ECS algorithm against four other competitive algorithms has also been shown.