Optimization of a Hybrid NLP Model for Multi-Aspect Sentiment Analysis of the Minister of Finance of the Republic of Indonesia

Abstract

The rapid growth of social media platforms has created extensive opportunities to analyze public opinion toward government policies using computational approaches. Public discussions related to fiscal governance, particularly those involving the Minister of Finance of the Republic of Indonesia, generate large volumes of textual data reflecting diverse societal perceptions. However, conventional sentiment analysis methods often fail to capture multidimensional opinions expressed across different policy aspects. Therefore, this study aims to optimize a hybrid Natural Language Processing (NLP) model for conducting multi-aspect sentiment analysis on public discourse collected from the X social media platform. The proposed framework integrates machine learning and deep learning techniques through a hybrid stacking approach combined with preprocessing optimization, class imbalance handling, and hyperparameter tuning. The dataset underwent text normalization, aspect identification, and sentiment labeling processes before model training and evaluation. Experimental results indicate that the hybrid NLP model achieved stable predictive performance with an accuracy exceeding 94%, demonstrating improved robustness in handling informal Indonesian social media language. Multi-aspect analysis reveals that public sentiment toward fiscal governance tends to be predominantly neutral, indicating analytical rather than emotionally polarized discussions across taxation, budgeting, and economic management aspects. The findings confirm that hybrid NLP combined with aspect-based sentiment analysis provides deeper interpretative capability compared to conventional sentiment classification approaches. This study contributes to the advancement of sentiment analysis research in low-resource language environments and offers a data-driven framework that can support digital governance evaluation and evidence-based policymaking through intelligent public opinion monitoring systems.