http://eprints.unpak.ac.id/id/eprint/10657 2026-04-18T06:05:36Z info:eu-repo/semantics/descriptiveMetadata http://eprints.unpak.ac.id/10657/ Sentiment Analysis of the RUU Perampasan Aset Using a Comparative Approach on Bidirectional Encoder Representations from Transformers (BERT) Models Alindiana, Resi Ilmu Komputer Sentiment Analysis of the RUU Perampasan Aset Using a Comparative Approach on Bidirectional Encoder Representations from Transformers (BERT) Models Resi Alindiana1*, Arie Qur'ania2 , Heri Bambang Santoso3 1 Department of Computer Science, Faculty of Mathematics and Natural Sciences, Pakuan University, Bogor, Indonesia 2 Department of Computer Science, Faculty of Mathematics and Natural Sciences, Pakuan University, Bogor, Indonesia 3 Department of Computer Science, Faculty of Mathematics and Natural Sciences, Pakuan University, Bogor, Indonesia * Correspondence: [065121081@student.unpak.ac.id] Abstract: The RUU Perampasan Aset has existed since 2003 but has not yet been enacted and continues to attract public attention, particularly on the social media platform X (Twitter), alongside the increasing discussion of corruption cases. Therefore, this study aims to analyze public perception of the RUU Perampasan Aset from a linguistic perspective. Sentiment analysis is conducted by comparing three BERT-based models—mBERT, IndoBERT, and IndoBERTweet which are capable of understanding bidirectional textual context. Sentiment classification is divided into positive and negative categories. Positive sentiment represents supportive opinions expressed in polite and ethical language, while negative sentiment represents supportive opinions conveyed in a sarcastic and pessimistic manner. The research process follows the Knowledge Discovery in Databases (KDD) methodology. Evaluation results using variations of epochs and batch sizes show that IndoBERT achieves the best performance with an accuracy of 92%, followed by mBERT at 91% and IndoBERTweet at 90%. These findings indicate performance differences among the three models, suggesting that model effectiveness is influenced not only by architectural design but also by dataset characteristics . Keywords: RUU Perampasan Aset, Sentiment Analysis, X (Twitter), mBERT, IndoBERT, IndoBERTweet. 2025-06-16 Thesis NonPeerReviewed Alindiana, Resi (2025) Sentiment Analysis of the RUU Perampasan Aset Using a Comparative Approach on Bidirectional Encoder Representations from Transformers (BERT) Models. Skripsi thesis, Universitas Pakuan. info:eu-repo/semantics/humanStartPage