<![CDATA[This study aims to conduct a sentiment analysis of user reviews of the Honor of Kings game on Google PlayStore using the Naive Bayes (NB) algorithm and Support Vector Machine (SVM) as a machine learning approach. The research gap raised in this study lies in the lack of comparative studies that quantitatively measure the performance of the two classic algorithms on Indonesian-language mobile game review data, as well as the absence of numerical mapping of sentiment distribution that describes user perceptions proportionally. The dataset used consisted of 1000 reviews, which after the manual labeling process was divided into 780 positive reviews (52%), 540 negative reviews (36%), and 180 neutral reviews (12%). The quantitative objective of this study was to measure and compare the levels of accuracy, precision, recall, F1-score, and AUC of the two models to determine the most effective algorithm in classifying user opinions. The test results showed that the SVM model produced an accuracy of 75.3% with an AUC value of 0.82, while the NB model obtained an accuracy of 71.1% with an AUC of 0.78. Based on the confusion matrix, SVM is able to reduce misclassification of negative and neutral sentiments, which are generally difficult to distinguish due to the distribution of sparse text features. Scientifically, this study contributes by showing that SVM is more optimal than NB in handling unbalanced review data, and confirms the importance of feature weighting and AUC validation as indicators of model reliability. Practically, the results of this study can be used by the developers of Honor of Kings to evaluate aspects of the user experience based on the sentiment patterns identified, especially in improving server stability, character balance, and player satisfaction.Keywords— Sentiment Analysis, Naive Bayes, Support Vector Machine, Honor of Kings, Google PlayStore. ]]>
