Cross-Sector Sentiment and Topic Analysis of Indonesian Digital Platform Applications: A Text Mining Study of Google Play Store Reviews

Abstract

The rapid expansion of Indonesia's digital economy has produced a large, continuously generated body of user feedback on mobile application stores, yet most prior sentiment-analysis studies examine a single application or a single sector in isolation, leaving cross-sector comparison largely unexplored. This study applies text mining to 841,664 Google Play Store reviews collected from eight widely used Indonesian digital platform applications spanning five sectors: fintech e-wallets (DANA, OVO), e-commerce (Blibli, Bukalapak, Tokopedia), on-demand services (Gojek), education technology (Ruang Guru), and travel (Traveloka). A rating-derived sentiment label (negative, neutral, positive) was validated using a TF-IDF and Logistic Regression classifier, which achieved 85.8% accuracy and a macro F1-score of 0.858 on held-out data. Non-negative matrix factorization (NMF) topic modeling was applied to negative reviews within each sector to surface dominant pain-point themes, and a Kruskal-Wallis test compared rating distributions across sectors. Results show statistically significant differences in user satisfaction across sectors (H = 160,561.5, p < .001, ε² = 0.191), with the fintech e-wallet sector exhibiting markedly lower satisfaction (63.4% negative reviews, mean rating = 2.35) than education technology (9.6% negative, mean rating = 4.51). Topic modeling reveals that e-wallet dissatisfaction concentrates on balance/fund-safety trust concerns and identity-verification (KYC) upgrade friction, distinguishing it from complaints in other sectors that center on pricing and service delivery. The findings extend Technology Acceptance Model/UTAUT-based fintech adoption research with large-scale textual evidence and offer practical guidance for prioritizing fintech user-experience improvements around trust and verification processes.