Hybrid PSO-GSA based approach for feature selection

With the development and widespread use of social networks among people, high-volume data is produced and the analysis of this data can be useful in many areas, including people's daily lives. Classification of this volume of data using traditional methods is a very difficult, time-consuming, and low-accuracy task, therefore, using sentiment analysis techniques, people's opinions can be effectively summarized and categorized. To this end, we propose an algorithm that combines Particle Swarm Optimization (PSO) and Gravitational Search Algorithm (GSA). The reason for combining the two algorithms is that the GSA has a good ability to search overall, but in the last iterations, it has a low speed in exploiting the search space. Since the PSO algorithm has a special ability to exploit the search space, this algorithm is used in the exploitation phase to solve the problem. The accuracy obtained from our proposed algorithm (PSO-GSA) shows an improvement in the accuracy of the GSA algorithm.