Island-based Cuckoo Search with elite opposition-based learning and multiple mutation methods for solving optimization problems

Abstract

<p>The island Cuckoo Search (<i>i</i>CSPM) algorithm is a variation of Cuckoo Search that uses the island model and highly disruptive polynomial mutation to solve optimization problems. This article introduces an improved <i>i</i>CSPM algorithm called <i>i</i>CSPM with elite opposition-based learning and multiple mutation methods (<i>i</i>CSPM2). <i>i</i>CSPM2 has three main characteristics. Firstly, it separates candidate solutions into several islands (sub-populations) and then divides the islands among four improved Cuckoo Search algorithms: Cuckoo Search via Lévy flights, Cuckoo Search with highly disruptive polynomial mutation, Cuckoo Search with Jaya mutation and Cuckoo Search with pitch adjustment mutation. Secondly, it uses elite opposition-based learning to improve its convergence rate and exploration ability. Finally, it makes continuous candidate solutions discrete using the smallest position value method. A set of 15 popular benchmark functions indicate <i>i</i>CSPM2 performs better than <i>i</i>CSPM. However, based on sensitivity analysis of both algorithms, convergence behavior seems sensitive to island model parameters. Further, the single-objective IEEE-CEC 2014 functions were used to evaluate and compare the performance of <i>i</i>CSPM2 to four well-known swarm optimization algorithms: distributed grey wolf optimizer, distributed adaptive differential evolution with linear population size reduction evolution, memory-based hybrid dragonfly algorithm and fireworks algorithm with differential mutation. Experimental and statistical results suggest <i>i</i>CSPM2 has better performance than the four other algorithms. <i>i</i>CSPM2's performance was also shown to be favorable compared to two powerful discrete optimization algorithms (generalized accelerations for insertion-based heuristics and memetic algorithm with novel semi-constructive crossover and mutation operators) using a set of Taillard's benchmark instances for the permutation flow shop scheduling problem.