Sentiment Analysis of Distance Learning During the Covid-19 Pandemic Using Particle Swarm Optimization (PSO) and Naïve Bayes Methods

The implementation of Distance Learning (PJJ) during the COVID-19 pandemic has sparked various responses from the public on social media, particularly on Twitter, with positive and negative sentiments that are difficult to sort due to the diverse use of language by users. This study aims to conduct sentiment analysis to extract information related to the PJJ policy using the Naïve Bayes method and Particle Swarm Optimization (PSO) for feature selection to improve accuracy. The data used consists of $\mathbf{1 0, 0 0 0}$ tweets collected from January to March 2022, which, after cleaning and labeling, resulted in $\mathbf{1, 2 1 8}$ tweets analyzed. The methods applied include tokenization, case folding, stemming, stopword removal, and data balancing using SMOTE. The results show that the use of PSO enhances the performance of the Naïve Bayes model, with an average accuracy of $\mathbf{9 7 . 2 9 \%}$. The key contribution of this research is the application of PSO to improve accuracy in sentiment analysis. Although the effect of PSO is not substantial, it still improves the overall model performance.