Public sentiment analysis of government subsidy policies on Twitter using the Naïve Bayes classifier

Abstract

Government subsidy policies are one of the important instruments in
maintaining economic stability and improving public welfare; however,
they often generate diverse responses in the public sphere. These differing
perspectives arise because subsidy policies directly affect the social and
economic lives of the community. In the digital era, social media
particularly the Twitter platform has become a medium for the public to
express opinions, criticisms, and information in real time regarding such
policies. This study aims to analyze public sentiment toward government
subsidy policies on the Twitter platform using the Naïve Bayes Classifier
method with text preprocessing stages. The research data consist of
Indonesian-language tweets collected from the Twitter platform during a
specific period in 2024. The text preprocessing stages include case folding,
tokenization, filtering/stopword removal, and stemming to eliminate
irrelevant words before the sentiment classification process into positive,
negative, and neutral categories. The results show that out of 87 analyzed
tweets, neutral sentiment dominates with a percentage of 66.67%,
followed by positive sentiment at 19.54% and negative sentiment at
13.79%. The dominance of neutral sentiment indicates that most tweets
are informational in nature, while expressed opinions tend to be more
positive than negative. Overall, these findings demonstrate that the Naïve
Bayes Classifier method is able to provide an objective overview of public
sentiment trends toward government subsidy policies on the Twitter
platform and can be utilized as a references for policy evaluation.