<![CDATA[TikTok has emerged as one of the fastest-growing social media platforms in 2025, especially among the younger generation. Beyond being a space for creative content sharing, TikTok has also become a vital platform for the exchange of public opinion, primarily through user comments. As user engagement intensifies, sentiment analysis on TikTok comments becomes increasingly essential to understanding public perception of various issues, trends, public figures, and brands. This study aims to analyze sentiment in TikTok user comments using machine learning classification algorithms. The research compares the performance of three widely used algorithms in text classification: Naive Bayes, Support Vector Machine (SVM), and Random Forest. A dataset of 5,000 public TikTok comments was collected through web scraping of trending videos from January to March 2025. The comments, written in Indonesian, underwent several text pre-processing steps, including tokenization, stopword removal, and stemming, to normalize the data. The TF-IDF method was then applied to extract numerical features from the textual data. A stratified data split was used to divide the dataset into training (80%) and testing (20%) subsets, ensuring balanced sentiment class distribution. Performance evaluation was conducted using accuracy, precision, recall, and F1-score metrics. Among the tested models, SVM achieved the highest accuracy of 89.7%, outperforming Naive Bayes and Random Forest across all metrics. These results indicate that SVM is particularly well-suited for classifying short, informal text such as TikTok comments. The findings contribute to the advancement of sentiment analysis in social media environments, specifically for Indonesian language data on TikTok. Moreover, the study provides valuable insights for industry stakeholders, marketers, and academic researchers seeking to implement data-driven public opinion analysis using machine learning techniques on emerging social media platforms.]]>
