Sentiment Analysis of Google Play Store Reviews on the Top Three Digital Banks in Indonesia Using Bidirectional Encoder Representations from Transformers (BERT)

This study aims to analyze user sentiment toward the top three digital banks in Indonesia—Seabank, Bank Jago, and Blu by BCA—by employing the Bidirectional Encoder Representations from Transformers (BERT) model. A total of 5.174 user reviews were collected from the Google Play Store and classified into positive and negative sentiments using a fine-tuned IndoBERT model. The reviews were preprocessed through tokenization, stemming, stopword removal, and normalization. BERT was selected for its proven capability to capture contextual relationships and semantic nuances in natural language. The model was evaluated using accuracy, precision, recall, and F1-score metrics. The sentiment classification achieved high accuracy, with results of 94.04% for Seabank, 92.39% for Bank Jago, and 90.60% for Blu. Sentiment distribution analysis revealed that Seabank had a predominantly positive sentiment (58.7%), whereas Bank Jago and Blu showed predominantly negative sentiments, at 54.6% and 55.7%, respectively. These results indicate that Seabank users are generally satisfied with the service, while Bank Jago and Blu may need to address service-related issues to enhance user satisfaction. Although sentiment polarity provides useful insights, it may not fully represent the complexity of user experiences. The insights derived from this research can assist digital banking application developers and stakeholders in identifying areas for improvement and enhancing the overall user experience in Indonesia's growing digital financial ecosystem.