GSA for machine learning problems: A comprehensive overview

The rapidly increasing data volume produced daily is encouraging to generate novel procedures for extracting suitable information from such data. Machine learning is an application of artificial intelligence which is employed to provide relevant knowledge extracted from data, due to such characteristics, the adoption of machine learning approaches is one of the most accepted alternatives for this purpose nowadays. On the other hand, many machine learning applications turn into complex tasks due to the nature of data and the procedure that these must be subjected to collecting appropriate information. Alternatively, metaheuristic techniques are optimization algorithms widely used for treating complex tasks efficiently. The Gravitational Search Algorithm (GSA) is an optimization method based on the Newtonian gravitational laws and the interaction of masses, this procedure has proved interesting results due to the employed operators for correct balancing the exploration and exploitation stages, avoiding the common flaws present in existing optimization techniques such as the premature convergence onto local minimal. In this study, a comprehensive overview of the GSA applied in several machine learning applications is carried out. (C) 2020 Elsevier Inc. All rights reserved.