<![CDATA[The increasing intensity of user interaction on the TikTok platform makes the comment section vulnerable to the emergence of rude comments, impolite speech, and negative verbal expressions that can reduce the quality of digital communication. The characteristics of TikTok language, which is informal, concise, and rich in slang variations and non-standard spelling, present challenges in the process of automatically identifying rude comments, especially in the Indonesian context. This study aims to develop and evaluate a binary classification model capable of distinguishing rude and non-rude comments on the TikTok platform using a text-based machine learning approach. The research method began with the collection of 650 Indonesian-language public comments from TikTok, which were then manually annotated into two classes: rude and non-rude comments. The labeled data were processed through preprocessing stages including text cleaning, case folding, slang normalization, repeated character reduction, tokenization, and stopword removal. Feature representation was carried out using the Term Frequency–Inverse Document Frequency (TF-IDF) method with a combination of unigrams and bigrams, while the classification process used the Logistic Regression algorithm. The data were divided into training data and test data with a ratio of 80:20. The analysis techniques used included evaluating model performance using accuracy, precision, recall, and F1-score metrics. The results showed that the model achieved an accuracy of 87.4%, with precision, recall, and F1-score values of 0.87 each, indicating good and balanced classification performance across both classes. These findings indicate that the combination of TF-IDF and Logistic Regression is effective as a baseline in classifying abusive Indonesian comments on the TikTok platform.]]>
