Sentiment Analysis of Positive and Negative Comments on the Rise of the Dollar: A Case Study from TikTok Using the Naive Bayes Method

Abstract

This study investigates public sentiment toward the appreciation of the United States dollar against the Indonesian rupiah through comments collected from TikTok. The research applies an empirical machine learning framework that integrates social media based economic discourse analysis with the Multinomial Naive Bayes classification algorithm. Data were processed through a structured preprocessing pipeline consisting of text normalization, removal of irrelevant characters, and feature extraction using the Bag of Words approach. Sentiment categories were assigned through a lexicon guided labeling procedure and subsequently classified into positive and negative classes. Model validation was conducted using an independent testing dataset and evaluated through accuracy, precision, recall, F1 score, and confusion matrix analysis. The findings indicate that the classifier achieved a high level of predictive performance, demonstrating the computational efficiency of Naive Bayes for large scale textual data. At the same time, the evaluation reveals methodological challenges associated with class imbalance and limited sentiment coverage arising from rule based labeling. The study highlights the value of social media analytics for monitoring economic perceptions and contributes to the development of sentiment intelligence frameworks capable of supporting real time observation of macroeconomic discourse in digital environments.