<![CDATA[The Free Nutritious Meal Program (Program Makan Bergizi Gratis, MBG) is a national policy aimed at improving public nutritional quality, particularly among school-aged children. Despite its strategic objectives, the program’s implementation has generated diverse public responses, widely expressed through social media platforms. This study aims to analyze public sentiment toward the Free Nutritious Meal Program using the Claude Sonnet 4.5 model with a zero-shot classification approach. The study was conducted online using YouTube user comments on videos discussing the MBG program as the data source. Data were collected through the YouTube Data API between January 5 and January 20, 2026, yielding a total of 5,036 comments. After preprocessing, 4,737 clean comments were retained for analysis. Sentiment classification was performed without model retraining by leveraging the contextual understanding capabilities of Claude Sonnet 4.5. Model performance was evaluated using a Confusion Matrix by comparing automatic classification results with manual labels on 20% of the data as an evaluation sample. The results indicate that relevance classification achieved an accuracy of 97.89%, while sentiment classification reached an accuracy of 94.60%. Sentiment distribution was dominated by negative sentiment at 57.8%, followed by neutral sentiment at 21.5% and positive sentiment at 20.7%. This study contributes by proposing a Large Language Model–based framework for public policy sentiment analysis using Claude Sonnet 4.5 with a zero-shot classification approach, enabling accurate analysis of Indonesian-language public opinion without reliance on labeled training data.]]>
