An Improved Elk Herd Optimiser (IEHO)

The Elk Herd Optimiser (EHO) is a novel metaheuristic algorithm inspired by the reproductive behaviour of elk herds. However, it suffers from insufficient convergence accuracy and population diversity. To address these issues, this study proposes an improved EHO (IEHO). A novel individual update strategy for the breeding phase is introduced to meet the requirements of convergence speed and diversity during evolution. A new population grouping strategy is also developed to achieve a dual balance between elite guidance and spatial distribution. Cauchy distribution sampling is used to generate learning weights, and population diversity is adopted to control the step size of movement, allowing real-time monitoring and supplementation of population diversity. A differentiated learning strategy based on fitness ranking divides individuals into high-quality and ordinary categories, implementing elite guidance and swarm intelligence learning, respectively. A hybrid evolutionary mechanism, integrating reverse learning driven by generalised opposition and perturbation based on the Cauchy distribution, substantially strengthens the algorithm’s resistance to entrapment in local optima. Moreover, dimension-masked crossover operations are introduced to greatly optimise the efficiency of population information sharing. Finally, through comparative experiments to verify the overall performance of IEHO on the CEC2017 test suite, compared with other competing algorithms, IEHO achieves the highest number of optimal solutions across multiple test functions.