<![CDATA[User reviews on the Google Play Store contain a wide range of opinions regarding the experience of using the Roblox game, both positive and negative. The large number of reviews makes manual sentiment identification inefficient, necessitating a sentiment analysis method capable of processing reviews systematically. This study aims to classify the sentiment of Roblox game reviews using TF-IDF Weighted Word2Vec as the feature representation and Extreme Gradient Boosting (XGBoost) as the classification algorithm. A total of 1,179 reviews were collected via web scraping during November 2025, then manually labeled and validated by an Indonesian language expert. The research stages included preprocessing—such as case folding, cleaning, tokenization, stopword removal, and stemming—followed by feature representation using TF-IDF Weighted Word2Vec with 200-dimensional vectors. The dataset was split in an 80:20 ratio for training and testing, while model optimization was performed using Grid Search with 5-fold Cross Validation. The labeling results consisted of 786 positive reviews and 393 negative reviews. The best configuration yielded an average validation accuracy of 89.5%, while the test data achieved an accuracy of 0.89. The positive class achieved precision, recall, and F1-score of 0.91, 0.92, and 0.91, respectively, while the negative class achieved 0.83, 0.82, and 0.83. These results indicate that the combination of TF-IDF Weighted Word2Vec and XGBoost is capable of effectively classifying the sentiment of Roblox game reviews.]]>
