Cuckoo search algorithm with ensemble strategy and the application in chaotic time series prediction

Abstract Cuckoo search (CS) algorithm is a simple and effective search technique. However, CS algorithm may suffer from premature convergence as the complexity of the problem increases. To address this challenge, a cuckoo search algorithm with ensemble strategy, called CSES, is presented to strengthen the convergence performance. Specifically, three new search strategies with diverse properties are designed to well balance the trade-off between global exploration and local exploitation. After that, according to the idea of selective ensemble, a priority roulette method is employed to select the appropriate search strategy at different stages of the evolution process, so as to produce more promising results. To investigate the comprehensive performance of CSES algorithm, extensive experiments are carried out on 53 benchmark functions and three chaotic time series prediction problems. Simulation results illustrate that the proposed CSES is superior to six recently developed CS variants and several other advanced evolutionary algorithms.