Feature Selection based on Particle Swarm Optimization Algorithm for Sentiment Analysis Classification

Online media serve as a potential secondary data source for studies on sentiment analysis. The current conditions of the data sources are very different, and it offers a variety of writing systems. Therefore, the results of accuracy in sentiment analysis are very important. An improved approach was proposed to increase the sentiment analysis accuracy based on text pre-processing and Naïve Bayes Classifier algorithm hybrid with Particle Swarm Optimization (NBC-PSO). Furthermore, the proposed algorithm solves the complex background problems about noise data and feature selection that affect the classification performance on sentiment analysis. This proceeded with the classification of positive or negative sentiments on these texts using NBC. Subsequently, the feature selection based on PSO was created to improve the accuracy. The experimental results showed that the proposed approach has a significant effect on sentiment score and polarity detection.