A ranking-based adaptive cuckoo search algorithm for unconstrained optimization

Cuckoo search (CS) has been proven to be one of the most efficient metaheuristic algorithms in solving global optimization problems. However, it suffers from a slow convergence speed and premature convergence, especially when the complexity of the problem increases. To address these shortcomings, a ranking-based adaptive cuckoo search algorithm, called RACS, is proposed in this paper. Specifically, a novel ranking-based mutation strategy is designed at first, which is inspired by the natural phenomenon that good species or individuals always contain good information and thus have better odds of guiding others. In the proposed ranking-based mutation strategy, the global search equation is modified in combination with a ranking-based vector selection method, where some of the parent vectors are proportionally selected according to their rankings. The higher ranking a parent vector obtains, the more opportunity it will be chosen. Secondly, a crossover operation with parameter adaptation is employed after the Levy flights random walk to preserve some good elements of the current solutions from being changed. Furthermore, a replacement strategy is designed to update the solutions not improved through pre-determined cycles by exploiting the beneficial information from the discarded solutions saved in the external archive. To evaluate the comprehensive performance of RACS, extensive experiments are conducted on three well-known test suites and an application problem of identifying unknown parameters of fractional-order nonlinear systems. Simulation results demonstrate that the presented strategies bring a significant improvement in effectiveness and efficiency on CS. Besides, RACS is verified to be superior or at least comparable to other CS variants and state-of-the-art algorithms on most of the benchmark problems, and thus, can be regarded as a useful and promising technique for solving real-world complex optimization problems.