Sentiment Analysis of TikTok Comments on Free Nutritious Meals Using Naïve Bayes

Abstract

Public opinions on various issues, including government programs, are increasingly expressed through social media platforms. TikTok, as one of the most active social media platforms, allows users to share comments, criticism, and support regarding the Free Nutritious Meals Program. This study examines TikTok remarks about the Free Nutritious Meals Initiative and classifies them into sentiment categories using the Naïve Bayes algorithm. A total of 2,094 comments were obtained through a crawling process and then manually grouped into positive, negative, and neutral classes. The labeling process was conducted manually based on predefined sentiment guidelines and was reviewed to improve label consistency. Before being used in the classification stage, the comments were cleaned and standardized through several preprocessing steps, such as cleaning, case folding, tokenization, stopword removal, and stemming. After preprocessing, the text data were transformed into numerical representations using the Term Frequency–Inverse Document Frequency method, and the classification process was carried out with the Naïve Bayes algorithm. The results showed that neutral sentiment had the highest proportion at 45.9%, followed by negative sentiment at 33.9% and positive sentiment at 20.2%. The model achieved an accuracy of 73.75%, with precision of 68%, recall of 69%, and F1-score of 69%. These findings indicate that Naïve Bayes is able to classify TikTok user sentiment toward the Free Nutritious Meals Program with reasonably good performance, although informal language, ambiguous expressions, and sarcasm remain challenges in the classification process.