An escalated convergent firefly algorithm

Abstract

Firefly algorithm is a powerfulalgorithm, however, it may show inferior convergence rate towards global optimum because the optimization process of the firefly algorithm depends upon on a random quantity that facilitates the fireflies to explore the search space at the cost of the exploitation. As a result, to improve the performance of firefly algorithm in terms of convergence rate even as its ability to exploit the solutions is preserved, firefly algorithm is altered using two different approaches in this research work. In the first approach, a local search strategy, i.e., classical unidimensional local search is employed in the firefly algorithm in order to improve its exploitation capability. In the second approach, a new solution search strategy, i.e., stochastic diffusion scout search is integrated with the searching phase of the firefly algorithm in order to increase the probability of inferior solutions to improve themselves. Both the approaches are evaluated on various benchmark problems having different complexities and their performances are compared with artificial bee colony and classical firefly algorithm. Results prove the capability of the proposed approaches in finding the global optimum in the search space and avoid the local optima at the same time.