Text Mining Analysis of SeaBank Reviews: Review Distribution, Frequent Words, and Topic Modeling Using LDA

Abstract

This study analyzes SeaBank user reviews using text mining techniques to identify dominant user perceptions toward the mobile banking application. The dataset was collected from Google Play Store reviews posted between January 1 and July 10, 2025, resulting in 90,814 filtered reviews after preprocessing. The analysis includes review distribution, word frequency analysis, and topic modeling using Latent Dirichlet Allocation (LDA). The results show that SeaBank reviews are predominantly positive, with frequently occurring words such as mudah (easy), bagus (good), and cepat (fast), indicating that usability, efficiency, and transaction performance are central to user experience. The LDA model was evaluated using Coherence Score and Log Perplexity, achieving a Coherence Score of 0.5787 and a Log Perplexity of -5.6624. Based on the highest-weighted keywords, four latent topics were identified and manually labeled as Usability, Features, Benefits, and Satisfaction. These findings suggest that user perception of SeaBank is shaped not only by the availability of banking services, but also by the application’s usability, service efficiency, and overall service quality.