<![CDATA[Sentiment analysis is used to identify and classify opinions expressed in text. In GTA V Roleplay player reviews, high-dimensional textual features and noisy data can reduce classification Accuracy and increase computational complexity. This study addresses these challenges by optimizing the Multinomial Naïve Bayes classifier using Chi-Square and Information Gain Feature Selection techniques. The novelty of this research lies in evaluating these methods on Indonesian-language GTA V Roleplay reviews, which contain informal language and gaming-specific vocabulary, a domain that has received limited attention in previous Indonesian studies. The dataset consists of 604 preprocessed reviews represented using TF-IDF and evaluated through 10-fold stratified cross-validation. The results show that Chi-Square achieved the best performance, with 83% Accuracy, 81% Precision, 96% recall, and an F1-score of 88%, outperforming Information Gain in recognizing negative sentiment. These findings provide useful insights for GTA V Roleplay server developers and contribute to the development of Indonesian-language sentiment analysis in the online gaming domain.]]>
Copyrights © 2026
