Sentiment Analysis on Work from Home Policy Using Naïve Bayes Method and Particle Swarm Optimization

Abstract

At the beginning of 2020, the world was shocked by the coronavirus, which spread rapidly in various countries, one of which was Indonesia. So that the government implemented the Work from Home policy to suppress the spread of Covid-19. This has resulted in many people writing their opinions on the Twitter social media platform and reaping many pros and cons of the community from all aspects. The data source used in this study came from tweets with keywords related to work from home. Several previous studies in this field have not implemented feature selection for sentiment analysis, although the method used is not optimal. So that the contribution in this study is to classify public opinion into positive and negative using sentiment analysis and implement PSO for feature selection and Naïve Bayes for classifiers in building sentiment analysis models. The results showed that the best accuracy was 81% in the classification using Naive Bayes and 86% in the classification using naive Bayes based on PSO through a comparison of 90% training data and 10% test data. With the addition of an accuracy of 5%, it can be concluded that the use of the Particle Swarm Optimization algorithm as a feature selection can help the classification process so that the results obtained are more effective than before.

Sentiment Analysis on Marketplace in Indonesia using Support Vector Machine and Naïve Bayes Method

“…This study evaluated the labeling of the three libraries with manual labeling. Because manual labeling on sentiment can result in better accuracy [47], [48]. Manual labeling is considered superior for the purpose of labeling evaluation in sentiment analysis because: -Human labeling is accurate and reliable as it can understand the context and meaning of the text.…”