IndoBERT-Based Sentiment Analysis of Electric Motorcycle Policy in Indonesia Using Instagram Data

This study aims to analyze public sentiment toward the procurement of electric motorcycles within the Nutritional Service Fulfillment Unit/ Satuan Pelayanan Pemenuhan Gizi (SPPG) program in Indonesia by utilizing data from Instagram. The approach employed is a deep learning-based sentiment analysis using the IndoBERT model, which has been fine-tuned to classify data into positive, negative, and neutral categories. The research stages include data collection, preprocessing, labeling, model development, and model evaluation using accuracy, precision, recall, and F1-score metrics. The results indicate that public sentiment is predominantly negative at 80%, followed by positive sentiment at 15% and neutral sentiment at 5%. Further analysis reveals that negative sentiment is primarily driven by issues related to budget prioritization, infrastructure readiness, and policy effectiveness, while positive sentiment is associated with environmental benefits and improved service distribution efficiency. The model evaluation demonstrates that IndoBERT achieves high performance, with an accuracy of 0.89, precision of 0.88, recall of 0.90, and F1-score of 0.89. These findings indicate that IndoBERT is effective in capturing the contextual nuances of the Indonesian language in unstructured social media data. This study contributes to the advancement of transformer-based sentiment analysis methods and provides data-driven insights to support more responsive and evidence-based policymaking.