<![CDATA[The rapid advancement of digital payment technologies has accelerated the widespread adoption of mobile wallet applications, making it increasingly important for service providers to understand user perceptions and experiences. User reviews published on mobile application platforms represent valuable sources of feedback that reflect satisfaction, complaints, and expectations regarding service performance. However, the large volume of textual reviews makes manual analysis inefficient and difficult to manage. This study aims to analyze user sentiment toward the DOKU e-wallet application by applying transformer-based natural language processing techniques. A total of 11,685 user reviews collected from mobile application platforms were analyzed using two transformer-based models. The analytical process followed a structured data mining approach, including data collection, preprocessing, model training, and evaluation using accuracy, precision, recall, and F1-score metrics. The results show that the IndoBERT model achieved an accuracy of 93.1%, while the GPT-3.5 Turbo model achieved 93.2%, indicating strong performance in sentiment classification tasks. In addition, the analysis identified several recurring issues reported by users, including account access problems, verification difficulties, transaction errors, and customer service responsiveness. This study contributes to the literature by providing a comparative evaluation of transformer-based models in the context of digital payment platforms, particularly within the Indonesian ecosystem.]]>
Copyrights © 2026
