Learning-based elephant herding optimization algorithm for solving numerical optimization problems

Abstract

The elephant herding optimization (EHO) is a recent swarm intelligence algorithm. This algorithm simulates the clan updating and separation behavior of elephants. The EHO method has been successfully deployed in various fields. However, a more reliable implementation of the standard EHO algorithm still requires improving the control and selection of the parameters, convergence speed, and efficiency of the optimal solutions. To cope with these issues, this study presents an improved EHO algorithm terms as IMEHO. The proposed IMEHO method uses a global velocity strategy and a novel learning strategy to update the velocity and position of the individuals. Furthermore, a new separation method is presented to keep the diversity of the population. An elitism strategy is also adopted to ensure that the fittest individuals are retained at the next generation. The influence of the parameters and strategies on the IMEHO algorithm is fully studied. The proposed method is tested on 30 benchmark functions from IEEE CEC 2014. The obtained results are compared with other eight metaheuristic algorithms and evaluated according to Friedman rank test. The results imply the superiority of the IMEHO algorithm to the standard EHO and other existing metaheuristic algorithms.