Implementation of BERTopic for Topic Modeling Analysis of the Free Nutritious Meal Program Based on YouTube Comments

Abstract

The Free Nutritious Meal Program (Makan Bergizi Gratis), represents a significant national effort aimed at mitigating stunting rates in Indonesia, having commenced its operations in January 2025. As the program progressed, public sentiment towards it evolved, resulting in a diverse array of opinions that were extensively debated on various social media platforms, notably YouTube. This study was conducted with the objective of examining the perceptions of the public regarding Makan Bergizi Gratis through a topic modeling methodology employing the BERTopic approach, which analyzed 19,843 comments from YouTube. The analytical framework entailed several stages, including data preprocessing, sentence-based embedding representation, dimensionality reduction via UMAP, clustering through HDBSCAN, and topic interpretation grounded in c-TF-IDF. The findings indicate that public commentary is categorizable into ten primary themes, encompassing issues such as the involvement of political figures, concerns over budget transparency, the program's educational benefits, and the need for equitable access in underserved regions. Evaluation results show that BERTopic outperformed the traditional LDA model, with a coherence score of 0.46 compared to 0.39 and topic diversity of 76 percent compared to 71 percent. This analysis reveals that public perception of Makan Bergizi Gratis is multifaceted, shaped by social experience, political context, and economic expectations. These insights may serve as a valuable foundation for a more comprehensive understanding of public opinion, thereby supporting more targeted and responsive policy development.The Free Nutritious Meal Program (Makan Bergizi Gratis), represents a significant national effort aimed at mitigating stunting rates in Indonesia, having commenced its operations in January 2025. As the program progressed, public sentiment towards it evolved, resulting in a diverse array of opinions that were extensively debated on various social media platforms, notably YouTube. This study was conducted with the objective of examining the perceptions of the public regarding Makan Bergizi Gratis through a topic modeling methodology employing the BERTopic approach, which analyzed 19,843 comments from YouTube. The analytical framework entailed several stages, including data preprocessing, sentence-based embedding representation, dimensionality reduction via UMAP, clustering through HDBSCAN, and topic interpretation grounded in c-TF-IDF. The findings indicate that public commentary is categorizable into ten primary themes, encompassing issues such as the involvement of political figures, concerns over budget transparency, the program's educational benefits, and the need for equitable access in underserved regions. Evaluation results show that BERTopic outperformed the traditional LDA model, with a coherence score of 0.46 compared to 0.39 and topic diversity of 76 percent compared to 71 percent. This analysis reveals that public perception of Makan Bergizi Gratis is multifaceted, shaped by social experience, political context, and economic expectations. These insights may serve as a valuable foundation for a more comprehensive understanding of public opinion, thereby supporting more targeted and responsive policy development.