An Extended Version of Flower Pollination Algorithm

Flower pollination algorithm (FPA) is a novel meta-heuristic algorithm inspired from the pollination of flowers. It has been applied to various fields of research and has proved its worth. But the algorithm suffers from certain limitations which confines its scope to different field of applications. In the present work, four modified versions of FPA have been proposed to improve exploration, searching accuracy, convergence speed and to maintain good balance between intensification and diversification. Sine Cosine operator has been added instead of Levy flight for improving the exploration and searching accuracy. Secondly, an improved probability switching has been used to balance the exploration and exploitation. The last modification uses improved distribution parameter for enhancing the local search phase. The results of proposed versions are tested on nineteen benchmark functions for different population sizes. The best modified version of algorithm has been named as linear-exponential flower pollination algorithm (LEFPA). The results of LEFPA are then compared with firefly algorithm, differential evolution, bat algorithm, bat flower pollination algorithm, Sine Cosine algorithm, particle swarm optimization, genetic algorithm, cuckoo search and FPA. Experimental results show that LEFPA algorithm is better when compared to other algorithms and can be considered as the potential candidate to become state of the art. Further, statistical testing has been done to prove the significance of the proposed approach.