Analyzing Public Sentiment Toward Indonesia's Free Nutritious Meals Program Using BERT on Social Media X: Implications for SDG 2, SDG 4, and Sustainable Waste Management

Abstract

This study examines public sentiment toward Indonesia's Free Nutritious Meals Program (MBG) within efforts to support Sustainable Development Goal (SDG) 2, SDG 4, and sustainable waste management through a Natural Language Processing approach. Data were collected from the social media platform X using web crawling, resulting in 2,613 posts that were preprocessed and labeled with VADER. The analysis was conducted using the Knowledge Discovery in Databases method, and classification was performed using the Bidirectional Encoder Representations from Transformers model. The results indicate that public sentiment is more negative (56.1 %) than positive (43.9%), with significant variation across provinces. Negative sentiment is mainly associated with food safety, transparency, and program implementation, while positive sentiment reflects public appreciation of the nutritional benefits for students. The BERT model demonstrates strong performance, achieving accuracy of 93.50%, precision of 93.56%, recall of 93.50%, and an F1-score of 93.48%, indicating its effectiveness in capturing complex linguistic contexts. These findings show that the program's success depends not only on its objectives but also on implementation quality and responsiveness to public perception. Therefore, policymakers should consider public opinion when improving the program. The findings also have implications for environmental governance because the program's operational scale poses significant food-waste risks.