Sentiment Analysis and Key User Experience Factors of the Ruang GTK App: Integrating the InSet Lexicon, Explainable Machine Learning, and Large Language Model-Based Thematic Analysis

Mulyadi S Mulyadi S (Universitas Stikubank)
Eka Ardhianto (Universitas Stikubank)

Formosa Journal of Computer and Information Science
Vol. 5 No. 2 (2026): August 2026

Ruang GTK is the Indonesian government's professional development app for teachers, but evidence on its user experience is limited. This study analyzed 10,000 recent Google Play reviews (November 2023–September 2026). Reviews were labeled using the InSet lexicon with additional negation rules and classified with TF-IDF using Linear SVM, Multinomial Naive Bayes, and Logistic Regression under repeated group k-fold cross-validation. Influential features were derived from model weights, and key factors were formulated through LLM-assisted thematic analysis verified by the researcher. Of the 9,239 reviews retained after preprocessing, 88.05% were positive, 7.07% neutral, and 4.88% negative. Linear SVM achieved the highest weighted F1 (0.908) and remained significantly better than both other models under the corrected resampled t-test. Five key factors were identified; the only negative factor concerned access problems after app updates.