<![CDATA[The popularity of online games continues to increase, including Free Fire, which has gained more than one billion downloads and millions of user reviews on the Google Play Store. However, the variation and inconsistency of user comments make manual sentiment evaluation difficult. This study aims to compare the performance of Support Vector Machine (SVM) and Naïve Bayes in classifying user review sentiment on the Free Fire game. A total of 535 Indonesian-language reviews were collected using web scraping and processed through text cleaning, case folding, normalization, stopword removal, and stemming. Sentiment labels were assigned manually based on review content. The dataset was divided into training and testing using a 70:30 ratio, and feature extraction used Term Frequency–Inverse Document Frequency (TF-IDF). Two scenarios were implemented: a baseline without class balancing and a scenario using Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance. Results show that SVM outperforms Naïve Bayes in both scenarios. In the baseline, SVM achieved 89.81% accuracy, while Naïve Bayes obtained 82.80%. After SMOTE, SVM improved to 91.08% accuracy and Naïve Bayes to 89.17%. These findings indicate that SVM, especially with SMOTE, provides a more effective and balanced performance for sentiment classification on Free Fire reviews. The study contributes to providing a more accurate understanding of user perception and strengthening model development for sentiment analysis on digital game applications.]]>
Copyrights © 2025
