Sentiment analysis using lexicon-based method with naive bayes classifier algorithm on #newnormal hashtag in twitter

• Published in: Journal of physics. Conference series Vol. 1918; no. 4; pp. 42155 - 42161
• Main Authors: Mustofa, R L, Prasetiyo, B
• Format: Journal Article
• Language: English
• Published: Bristol IOP Publishing 01.06.2021
• Subjects: Algorithms; Classification; Classifiers; COVID-19; Data mining; Dictionaries; Digital media; Feature extraction; Physics; Sentiment analysis; Social networks; Weight
• ISSN: 1742-6588; 1742-6596; 1742-6596
• DOI: 10.1088/1742-6596/1918/4/042155

Back In 2020, World Health Organization (WHO) has announced COVID-19 as a pandemic. From the many public responses, especially those on Twitter regarding the #newnormal campaign, a sentiment analysis process needs to be carried out to find out the perceptions that exist in society through social media. In this study, data were obtained through the crawling process on Twitter using the Twitter API. The method used in the sentiment analysis process is lexicon-based. The lexicon-based method works by labeling words containing sentiments based on a lexicon dictionary that already has weight on each word or doesn’t have weight on words in a lexicon dictionary. The classification results using lexicon-based are also used to make training data in the testing process using the naive Bayes classifier algorithm. In general, the research stages in this sentiment analysis include data crawling, text preprocessing, feature extractions, and the classification process. The sentiment analysis process results showed that the percentage of social media users on Twitter about #newnormal was 33.19% containing negative sentiments and 66.36% containing positive sentiments. Meanwhile, for testing the naive Bayes classifier algorithm in the sentiment analysis process got an accuracy of 79.72%.