Abstract—This article proposes a new gravitational coefficient function of the gravitational search algorithm (GSA). Since the function concerns to the performance of GSA, we investigate its characteristic which influences the algorithm on global search performance. The novel function is compared to a former function in literature on four benchmark functions which incorporated of both unimodal landscape functions and multimodal landscape functions. The experimental results show that the proposed gravitational coefficient function outperforms the conventional one. The proposed function also shows that it works well on multimodal landscape functions. By balancing between exploration phase and exploitation phase, the slow convergence rate is compensated by the better solutions.
Index Terms—GSA, metaheuristics, optimization problems, search performance.
Pattrawet Tharawetcharak is with Office of Learning Promotion and Provision Academic Services, Varaya Alongkorn Rajabhat University under The Royal Patronage, Pathumthani, 13180 Thailand and he is also with the Department of Industrial Engineering and Management, Faculty of Engineering and Industrial Technology, Silpakorn University, Nakhon Pathom, 73000 Thailand (e-mail: pattrawet@vru.ac.th).
Thanathorn Karot and Choosak Pornsing are with the Department of Industrial Engineering and Management, Faculty of Engineering and Industrial Technology, Silpakorn University, Nakhon Pathom, 73000 Thailand (e-mail: karot_t@su.ac.th, pornsing_c@su.ac.th).
Cite: Pattrawet Tharawetcharak, Thanathorn Karot, and Choosak Pornsing, "An Improved Gravitational Coefficient Function for Enhancing Gravitational Search Algorithm’s Performance," International Journal of Machine Learning and Computing vol. 9, no. 3, pp. 261-266, 2019.