Efficient boosted grey wolf optimizers for global search and kernel extreme learning machine training

Grey wolf optimizer (GWO) is a new nature-inspired algorithm that simulates the predatory behaviors of grey wolves in nature. The GWO mainly divides the whole hunting process into three stages: encircling, hunting, and attacking when they are nearby the prey. Since its introduction, the GWO has found its applications in a wide range of engineering and science fields. However, when tackling more complex optimization problems, especially the high dimensional and multimodal tasks, GWO may easily fall into the local optima or be unsuccessful in finding the global best. In addition, convergence behaviors may not be very satisfying. In this study, the performance of basic GWO is enhanced using effective exploratory and exploitative mechanisms such as random leaders, opposition-based learning, levy fight patterns, random spiral-form motions, and greedy selection. These concepts are utilized to improve the global exploration and local exploitation capacities of the conventional technique and deepen the searching advantages of GWO in dealing with more complex problems. Also, the proposed mechanisms can ameliorate the convergence inclinations and the quality of the solutions. In order to verify the efficacy of the proposed method, which is called OBLGWO; it is compared to a comprehensive set of the new and state-of-the-art optimizers on 23 benchmark test sets and 30 well-known CEC problems. Additionally, the proposed OBLGWO is also applied to the tuning of the key parameters of kernel extreme learning machine (KELM) in dealing with two real-world problems. The experimental results and analysis demonstrate that the proposed OBLGWO can significantly outperform GWO, previous enhanced GWO variants and some of the other well-established algorithms in terms of convergence speed and the quality of solutions. (C) 2019 Elsevier B.V. All rights reserved.