<![CDATA[The rapid development of mobile-based information technology has increased the relevance of sentiment analysis on user-generated reviews. This study applies the Support Vector Machine (SVM) method combined with TF-IDF feature extraction to classify user sentiments from WhatsApp and Telegram reviews on the Play Store. A total of 4,400 Indonesian-language reviews (2,200 per application), collected via web scraping from different time periods in 2024, were processed through standard text preprocessing techniques and transformed using TF-IDF with 1–2 n-grams. The SVM model with a linear kernel was trained on 80% of the data and tested on 20% using accuracy, precision, recall, and F1-score metrics. Results show that Telegram reviews achieved higher accuracy (83%) and F1-score (0.83) compared to WhatsApp (68% accuracy, 0.73 F1-score). Sentiment analysis revealed a positive sentiment dominance in WhatsApp (~60%) and negative sentiment in Telegram (~52%). These findings suggest that Telegram reviews tend to have more concise and structured language, contributing to better classification performance. The study confirms the effectiveness of the SVM–TF-IDF approach and recommends further research using advanced models and embeddings to handle the complexity of informal review language.]]>
Copyrights © 2025
