Public Sentiment Complexity Using Fuzzy Logic: A Case Study of Indonesia’s New Sovereign Wealth Fund
Muhammad Fahri, Mahir Pradana, Budi Rustandi Kartawinata · 2025
Danantara, introduced in 2025 as Indonesia’s new Sovereign Wealth Fund (SWF), sparked significant public discourse, particularly on TikTok, due to transparency and political affiliation issues. This reflects the growing importance of social media sentiment in assessing public trust in state institutions. This study aims to analyze public sentiment toward Danantara by integrating Support Vector Machine (SVM) and Fuzzy Logic, which address the limitations of conventional three-class sentiment models. Using the CRISP-DM framework, 6,578 TikTok comments were collected and preprocessed. SVM with an RBF kernel achieved $74 \%$ accuracy in three-class categories (Negative, Neutral, Positive). Fuzzy Logic was then applied to enhance sentiment granularity into five categories (Very Negative, Negative, Neutral, Positive, and Very Positive), resulting in an accuracy of $67 \%$. While the model effectively identified extreme sentiments (Very Positive and Very Negative), it struggled with Neutral and Negative classes due to class imbalance and semantic overlap. These findings underscore the complexity of sentiment analysis in digital discourse. Future work should explore contextual models like BERT and data balancing methods. The results can support data-driven policy response and early public feedback monitoring.