<![CDATA[User reviews provide important insights into the quality of digital banking applications; however, their large volume makes manual analysis inefficient. This study applies Aspect-Based Sentiment Analysis (ABSA) to examine user perceptions of the BYOND by BSI application based on three aspects: interface, features and performance, and services. Three classification algorithms were compared: Naïve Bayes, Support Vector Machine (SVM), and Random Forest, evaluated with accuracy, precision, recall, F1-score, and ROC-AUC. The results indicate that SVM and Naïve Bayes achieved the best performance, with an accuracy of 0.95 and an F1-score of 0.92, whereas Random Forest exhibited slightly lower performance with an F1-score of 0.89. Furthermore, sentiment analysis reveals the features and performance aspect exhibits the highest proportion of negative sentiment (39.6%), primarily associated with system reliability issues, login problems, transaction failures, and application instability. These findings demonstrate that ABSA can serve as an effective knowledge discovery approach for identifying critical functional issues and supporting data-driven prioritization in improving digital banking services, particularly within the context of sharia banking applications.]]>
Copyrights © 2026
