Comparison of Different Text Classification Methods for Free Meal Policy Sentiment in Indonesia

Abstract

The free meal policy is an important initiative to improve the nutrition of children under five and pregnant women and reduce social inequality. This policy supports low-income families by providing free food and milk in schools and Islamic boarding schools. On social media, especially platform X (Twitter), this policy sparked public discussion. This research aims to analyze sentiment regarding the free meal policy using Naive Bayes, SVM, and Decision Tree methods, as well as providing the effectiveness of classification algorithms in understanding public opinion. Of the 5,205 tweets analyzed, there were 4,735 positive tweets and 470 negative tweets. Applying Smote to this analysis provides significant results. SVM achieved 99% accuracy, Decision Tree also showed good performance with 98% accuracy. Meanwhile, Naive Bayes experienced an increase in accuracy of up to 91%, although it was still less than optimal in detecting negative sentiment compared to SVM and Decision Tree.