Public Sentiment Classification of Danantara in Social Media X Using Support Vector Machine and Random Forest

Abstract

The increasing use of social media as a platform for public discourse provides valuable data for understanding societal responses to national strategic policies. One prominent example is the establishment of Danantara (Daya Anagata Nusantara), a sovereign wealth fund launched by the Indonesian government in February 2025. This study aims to analyze public sentiment toward Danantara using Indonesian-language posts collected from social media platform X and to comparatively evaluate the performance of Support Vector Machine (SVM) and Random Forest (RF) algorithms. A dataset of 1,434 public tweets was collected through web scraping and processed using text preprocessing techniques, including cleaning, tokenization, stopword removal, stemming, and TF-IDF feature extraction. Sentiment labels were generated using an Indonesian RoBERTa model and validated by a linguistic expert. Class imbalance was addressed using the Synthetic Minority Oversampling Technique (SMOTE). Model performance was evaluated using 5-fold stratified cross-validation with accuracy, precision, recall, and F1-score metrics. Experimental results show that Random Forest achieved slightly superior performance, reaching an average accuracy of 91.47%, compared to 91.06% obtained by SVM. Confusion matrix analysis indicates that RF better distinguishes neutral sentiment, while SVM performs competitively in identifying strong sentiment polarity. This study contributes by providing the first empirical comparison of classical machine learning approaches for analyzing public sentiment toward Indonesia’s sovereign wealth fund discourse, offering methodological insights and practical implications for data-driven policy evaluation using social media analytics.