<![CDATA[User reviews on the Google Play Store are unstructured data that is valuable for evaluating the quality of public broadcasting streaming application services. This study compares the performance of Support Vector Machine (SVM) and Multinomial Naive Bayes in classifying sentiment from 2,006 user reviews of the TVRI Klik application collected through web scraping. The sentiment labels are determined based on star ratings, namely four and five as positive, three as neutral, and one and two as negative. Labeling results in an unbalanced distribution of classes with negative classes dominating the data. Pre-processing includes case folding, tokenization, stopword removal, and stemming using Sastrawi, followed by TF-IDF feature extraction. SMOTE is applied to the training data to balance all three sentiment classes. With an 80:20 data sharing scheme, SVM obtained the highest test data accuracy of 95.13%, while Naive Bayes reached 86.59%. Wordcloud shows that negative reviews are dominated by technical complaints, while positive reviews highlight the ease of access to national broadcasts. The findings show that SVM with SMOTE is more stable and accurate for monitoring public sentiment towards streaming applications of government broadcasting institutions.]]>
Copyrights © 2026
