Review Sentiment Classification and Feature Selection Using Hybridized Support Vector Machine

Abstract

Recently, researchers have become increasingly curious about using machine learning algorithms to solve complex real-world problems. Machine learning algorithms can be broadly classified as either supervised learning techniques or unsupervised learning techniques. The expansion of social networks has significantly increased the volume of user-generated data, including reviews, comments, and opinions from customers. Processing this much user-generated content is difficult and time consuming, despite the fact that it might be helpful for analysis and decision-making. Therefore, it is essential to develop an intelligent system which automatically mines such vast amounts of content and categorizes them for positive and negative features. Sentiment analysis is useful for automatically monitoring social media, describing the general sentiment or attitude that customers have toward a specific brand or business, and determining whether they are regarded favorably or unfavorably online. By utilizing the Particle Swarm Optimization (PSO) and gray wolf optimization (GWO) strategies, the support vector machine (SVM) was investigated in order to create a novel and efficient prediction system. Support vector machine (SVM) is used in this analysis together with Particle Swarm Optimization (PSO) to classify the data. The experiment's findings demonstrate that PSO improves SVM performance, whereas in the other approach, GWO was utilized to update population positions in the discrete search space, resulting in the best feature subset for SVM-based classification purposes. The experimental data from this work demonstrates that SVM, PSO-SVM, and GWO-SVM have accuracy values of 79.76%, 81.35, and 82.18%, respectively. This shows that the GWO-SVM outperforms the competition.