The swift adoption of mobile banking in Indonesia highlights the growing demand for secure and innovative digital financial services. PT Bank Central Asia Tbk (BCA) offers two primary applications, BCA Mobile and myBCA, catering to millions of users. Gaining insight into user perceptions is crucial for enhancing service quality and building trust. This research uses sentiment analysis and topic modeling on Google Play Store reviews for both applications to facilitate knowledge discovery. Reviews were labeled using IndoBERT, and seven classification models were assessed, including five machine learning methods and two deep learning techniques. The Gated Recurrent Unit (GRU) model demonstrated the highest performance, achieving an accuracy of 89.70%. In the realm of topic modeling, a comparison between Latent Dirichlet Allocation (LDA) and BERTopic revealed that BERTopic delivered the highest coherence score of 0.6244, identifying eight significant negative topics. The findings indicate that BCA Mobile users frequently reported issues such as login failures, unexplained balance deductions, and missing features, while myBCA users encountered problems like post-update errors, login difficulties, and challenges with face verification. This research aligns with Sustainable Development Goal (SDG) 9 by showing how knowledge discovery from user reviews can promote innovation and enhance resilient, user-centered digital banking infrastructures.