<![CDATA[The rapid advancement of digital technology has spurred the emergence of online philanthropy platforms like Kitabisa, which collect a large volume of user reviews. Reviews on the Google Play Store reflect both satisfaction levels and service issues, but their unstructured nature makes manual analysis difficult. This study evaluates user sentiment on the Kitabisa platform by comparing the Support Vector Machine (SVM) and Naive Bayes models. A dataset of 11,887 reviews was processed through preprocessing and word weighting using the TF-IDF approach. The evaluation results show that the Support Vector Machine outperformed Naive Bayes with an accuracy of 84.05% and an F1-score of 0.93, while Naive Bayes achieved an accuracy of 81.73% and an F1-score of 0.92. Theoretically, this study reinforces the literature regarding the superiority of Support Vector Machines for unstructured text data. Additionally, the results of this research produce an automated evaluation framework that can be used by application developers as a basis for improving service quality in accordance with user perceptions accurately.]]>
Copyrights © 2026
