Comparative Sentiment Analysis of Digital Wallet Applications in Indonesia Using Naïve Bayes

Abstract

The rapid growth of financial technology in Indonesia has led to widespread use of digital wallet applications such as OVO, DANA, GoPay, and ShopeePay. User-generated reviews on platforms like the Google Play Store offer valuable insights into customer satisfaction and application performance. This study conducts a comparative sentiment analysis of user reviews for four major Indonesian e-wallets using the Multinomial Naïve Bayes algorithm. A total of 401 Indonesian-language reviews were collected and labeled based on user ratings, with sentiment classified as positive or negative. The TF-IDF method was applied for feature extraction, and the model was evaluated using accuracy, precision, and recall metrics. Results show that ShopeePay achieved the highest classification accuracy (89%), followed by DANA and GoPay (80%), while OVO recorded lower performance due to more informal and ambiguous language. The model demonstrated strong precision for positive sentiment but low recall for negative sentiment (28%), indicating challenges in detecting minority-class feedback. Word cloud visualizations were used to highlight common keywords in each sentiment category. This study confirms that Naïve Bayes is an effective approach for classifying user sentiment in Indonesian-language app reviews, while also emphasizing the need for improved handling of class imbalance in future research. The findings provide practical insights for developers to enhance user experience based on data-driven sentiment patterns.