<![CDATA[The increasing use of social media has created opportunities to examine public opinion regarding government policies through sentiment analysis approaches. This study analyzes public sentiment toward the budget allocation of the Makan Bergizi Gratis (MBG) program in Indonesia using Twitter/X data and the IndoBERT model. Data were collected through keyword-based crawling techniques, resulting in 24,836 tweets, which were reduced to 18,742 tweets after preprocessing. The dataset was manually annotated into positive, negative, and neutral sentiment categories before being classified using IndoBERT. The evaluation results indicate that the proposed model achieved strong performance, with an accuracy of 92.41%, precision of 91.86%, recall of 92.03%, F1-score of 91.94%, and ROC-AUC value of 0.961. The findings reveal that negative sentiment dominated public discourse, primarily associated with concerns regarding fiscal sustainability, transparency, and policy priorities. Positive sentiment mainly emphasized nutritional benefits and educational welfare impacts. The study demonstrates that transformer-based NLP models are effective for analyzing Indonesian social media discourse and provides practical insights for evidence-based policymaking and digital governance evaluation.]]>
Copyrights © 2026
