<![CDATA[Digital transformation increases the demand for fast and responsive technology-based public services through mobile applications, including the PDAM Info application. User reviews provide important insights for improving service quality, but their large volume makes manual analysis inefficient, requiring text-based sentiment analysis using machine learning. Default machine learning parameters are often suboptimal; therefore, Random Search is applied to improve classification performance. This study analyzes user sentiment and examines the effect of Random Search on sentiment classification of the PDAM Info application. A total of 2,400 Google Play Store reviews were collected, resulting in 1,677 data after preprocessing, labeled using a lexicon-based approach, and represented using TF-IDF. Logistic Regression and Support Vector Machine were used for classification with Random Search for hyperparameter tuning. The results indicate that negative sentiment dominates user reviews, mainly related to service coverage and payment methods. Random Search improves classification performance, achieving 88% accuracy and 83% F1-score, particularly in predicting positive and neutral classes on imbalanced data. The contribution of this study provides insights into user perceptions for PDAM Info application developers and demonstrates that Random Search.]]>
Copyrights © 2026
