<![CDATA[The increase in Land and Building Tax for Rural and Urban Areas (PBB-P2) has generated diverse public reactions that are widely expressed through social media platforms. TikTok, characterized by high user engagement and informal communication patterns, provides a relevant medium for observing public sentiment toward fiscal policies. This study examines public sentiment regarding the PBB-P2 increase by applying a Support Vector Machine (SVM)–based classification approach. Two kernel configurations, Linear and Polynomial, are compared to identify differences in classification behavior when handling social media text data. User comments were collected automatically and processed through text cleaning, feature extraction, and sentiment labeling stages prior to model training. Model performance was analyzed using confusion matrix–based evaluation to observe prediction patterns across sentiment classes. The findings indicate noticeable differences in classification stability between the two kernel types, with the Linear kernel showing more consistent behavior when applied to imbalanced sentiment distributions. These results suggest that selecting a kernel aligned with the characteristics of textual data is an important consideration in social media–based sentiment analysis of public policy issues]]>
Copyrights © 2026
