Analysis of Cuckoo Search Efficiency

Optimizing the parameters of a bio-inspired algorithm is naturally the main path to improve its performance. The distribution used in the displacement and creation of new solutions is also a factor to consider when enhancing its capacity. In this work, the efficiency of cuckoo search (CS) and self-adaptive cuckoo search algorithms (SACS) is investigated through extensive experimentation in three problems: (1) benchmark function optimization, (2) wind energy forecasting and (3) data clustering. This paper examines the reasons why CS and SACS have presented better performance and convergence rate than other algorithms in the above optimization problems. The Levy probability distribution employed in the algorithms, the reproduction strategy determined by parameters N and p(a), and the reduced number of parameters to optimize are candidate hypotheses studied. It is seen how such factors influence the performance of the algorithms, showing the efficiency of cuckoo search is very much associated with the Levy distribution.