<![CDATA[The rapid advancement of digital technology has fostered the rise of various interactive online gaming platforms, with Roblox standing out as one of the most prominent. This platform allows users not only to play but also to design and share their own games. As the number of active users increases, the volume of reviews submitted on the Google Play Store also grows. These reviews contain valuable information but require sentiment analysis to automatically understand users’ opinions, satisfaction levels, and complaints. This research aims to conduct sentiment analysis on Roblox user reviews by comparing the performance of three machine learning algorithms—Naïve Bayes, Random Forest, and Decision Tree—to determine which yields the most optimal results. The study follows the Knowledge Discovery in Databases (KDD) framework, which includes several stages: selecting 5,000 reviews, performing text preprocessing (such as cleaning, case folding, tokenizing, normalization, stopword removal, stemming, and labeling), transforming data using word embedding, and evaluating model performance with metrics including Confusion Matrix, Accuracy, Precision, Recall, and F1-Score. The experimental findings indicate that the Decision Tree algorithm achieved the best performance, with an accuracy of 85%, precision of 0.847, recall of 0.850, and a weighted F1-score of 0.848. In contrast, Random Forest obtained an accuracy of 83.6% and a macro F1-score of 0.773, while Naïve Bayes recorded the lowest performance with 64.2% accuracy and a macro F1-score of 0.527. Overall, the Decision Tree algorithm demonstrated superior capability and balance in classifying positive, negative, and neutral sentiments in Roblox user reviews, showing more effective text pattern recognition compared to probabilistic-based methods.]]>
