Optimizing self-adaptive gender ratio of elephant search algorithm by min-max strategy

Abstract

Elephant Search Algorithm (ESA) is one of the contemporary metaheuristic search recently proposed. Its efficacy depends largely on the right choice of gender ratio that balances the proportion between the number of male and female elephants as search agents with different functions. The male elephants are responsible for global exploration, roaming to new dimensions of search space. The female elephants focus on doing local search, for finding the optimal solution. The value for this gender ratio however needs to be manually chosen in the original version of ESA. An automatic mechanism for finding the appropriate gender ratio ESA agents is proposed in this paper. A self-adaptive method guided by min-max strategy is used to search for the optimal gender ratio of ESA. The self-adaptive method is simulated on nine optimization testing functions with different dimensions. Compared with enumerated global-best ratio, the self-adaptive ratio obtained by our method can save 90% of computation time at the cost of 20% compromise in fitness value in most testing functions. Simulation results are also compared with classical meta-heuristic algorithms including PSO, Firefly and WSA. ESA's performance with min-max ratio is also comparable towards these algorithms.