<![CDATA[The rapid growth of the social media platform TikTok has introduced new challenges in content moderation, particularly in detecting offensive language that appears not only in text form but also in visual elements such as video thumbnails. This study aims to develop a classification model capable of detecting offensive content in TikTok thumbnails using the Support Vector Machine (SVM) algorithm. Data were collected through web scraping of 4,153 TikTok videos containing offensive elements, which were then processed and manually labeled into 24 classes of offensive words. The dataset was divided into training and testing sets with a ratio of 20:80. Model performance was evaluated using AUC, accuracy, precision, recall, F1-Score, and Matthews Correlation Coefficient (MCC). The results show that the SVM model achieved an AUC of 0.791, indicating a reasonably good ability to distinguish between classes. However, accuracy (0.340), precision (0.293), recall (0.340), F1-Score (0.298), and MCC (0.264) indicate that the classification performance remains low. These findings suggest the need to improve Preprocessing quality, select more representative visual features, and develop more advanced classification methods. This research contributes to expanding the detection approach of harmful content from text-based to visual-based domains and lays the groundwork for more comprehensive automated content moderation systems in the future.]]>
Copyrights © 2025
