LAUSR

This study proposes a modified Elephant Herding Optimization algorithm to enhance the capability of a classical algorithm for convalescent convergence rate and precision to solve global optimization problems. The proposed Improved Elephant Herding Optimization (IEHO) uses an opposition learning-based initialization to get a better initial population. A sine cosine-based clan updating operator updates the clan individuals towards or outwards their clan leaders. Levy flight distribution with step size controller is applied to perform a local and global search on newly updated positions. The separating operator is modified to maintain a balance between exploration and exploitation of the algorithm. In addition, an elitism strategy is introduced to retain the fittest individual in the consequent iterations. The effectiveness of IEHO is validated on 97 benchmark functions which include unimodal, multimodal, and CEC-BC-2017 functions. The performance of IEHO is compared to fourteen state-of-the-art algorithms along with the winner algorithm of CEC-BC-2017. Friedman's mean rank test shows the dominance of the proposed algorithm for unimodal and multimodal functions. The proposed IEHO algorithm secures the best rank for all 97 benchmark functions. Finally, the applicability of IEHO is shown on five real-world engineering design problems. Results have proven that IEHO performed superior or equivalent to the algorithms reported in the literature and evaluated in this work.