Topic Modeling Analysis of I.Saku on the Play Store using Latent Dirichlet Allocation

Abstract

The development of digital wallets in Indonesia shows a rapid growth trend; however, not all applications can maintain user loyalty, including i.Saku, which has received a low rating on the Google Play Store. This study aims to identify the key factors shaping user reviews of the i.Saku application and to formulate service improvement recommendations based on those reviews. The method used is topic modelling with the Latent Dirichlet Allocation (LDA) algorithm applied to 3,000 user reviews from scraping the Play Store. Evaluation was conducted using coherence scores to determine the optimal number of topics, resulting in four main themes: (1) balance issues and customer service response, (2) transaction convenience and core features, (3) PIN and account access problems, and (4) login and verification obstacles. The analysis reveals that most negative reviews are related to technical issues and customer service, while positive reviews are dominated by ease of transaction. This study provides valuable insights for i.Saku developers to prioritize service improvements based on dominant issues identified in user reviews.