Sentiment Analysis of Pertamax on Social Media and MyPertaminta Data Using the IndoBert Algorithm

The rapid growth of digital services in Indonesia has accelerated the adoption of online platforms for fuel distribution through the MyPertamina application developed by PT Pertamina. Public responses toward the application and related fuel distribution policies are widely expressed through social media and application reviews. This study aims to analyze public sentiment toward MyPertamina using multi-platform data collected from Google Play Store, Instagram, and Twitter. The research employed a Natural Language Processing approach using the Transformer-based IndoBERT model. The methodology included data collection, data integration, text preprocessing, sentiment labeling, model fine-tuning, performance evaluation, and result visualization. The collected textual data were classified into positive and negative sentiment categories to represent public opinion. Experimental results showed that IndoBERT achieved an accuracy of 92.04%, with balanced precision, recall, and F1-score values. These findings demonstrate that IndoBERT effectively handles unstructured and informal Indonesian text from multiple digital platforms. Overall, integrating multi-platform data with IndoBERT-based sentiment analysis provides comprehensive insights into public perceptions of MyPertamina and supports strategic decisions. Future studies should expand data sources, increase dataset size, compare additional Transformer models, and evaluate broader sentiment patterns across diverse digital environments effectively.

Keywords: IndoBert, Multi-Platform Data, MyPertamina, Natural language Processing, Sentiment Analysis