<![CDATA[This study aims to analyze public sentiment toward the Indonesia Investment Authority (Badan Pengelola Investasi – BPI) Danantara using artificial intelligence technology. Data was collected through crawling using an X API token, resulting in 4,269 tweets stored in CSV format, consisting of 15 columns including tweet text and user metadata. The data underwent a pre-processing stage, including text cleaning, case folding, and tokenization, to prepare it for analysis. Manual labeling was conducted to classify sentiment into three categories: positive (32%), negative (45%), and neutral (23%). Due to class imbalance, a data augmentation technique was applied, increasing the total number of records to 23,623. The IndoBERT-base model was employed using a transfer learning approach for three-class sentiment classification. After five training epochs, the model achieved an accuracy of 97.71%. Evaluation results demonstrate high computational efficiency, with the model capable of processing data quickly. This study highlights the importance of applying artificial intelligence technologies, particularly BERT-based language models, in sentiment analysis in the digital era.]]>
Copyrights © 2025
