Application of Naive Bayes Algorithm for Analysis of User Reviews on Mobile Legends Game: Bang Bang

Mobile Legends: Bang Bang is a highly popular MOBA game, especially among students, which generates a large volume of user reviews on the Google Play Store. These reviews provide a valuable data source for understanding user sentiment. This study conducts sentiment analysis on user reviews using three variants of the Naïve Bayes algorithm: BernoulliNB, GaussianNB, and MultinomialNB. From an initial 5,000 reviews collected via web scraping using Python, 4,428 reviews were used after neutral reviews were removed to focus solely on positive and negative sentiments.The preprocessing steps included case folding, word normalization, tokenization, stopword removal, and stemming. Sentiment labeling was carried out using a lexicon-based approach, comparing the frequency of positive and negative words in each review. The dataset was split in an 80:20 ratio for training and testing.The results show that MultinomialNB achieved the highest accuracy at 75%, followed by BernoulliNB with 74%, and GaussianNB with 50%. MultinomialNB demonstrated superior performance in detecting positive sentiments, while BernoulliNB offered more balanced results. GaussianNB performed poorly due to its assumption of normally distributed continuous data, which is unsuitable for text classification. This study concludes that Multinomial Naïve Bayes is the most effective model for sentiment analysis of user reviews when working with word frequency-based representations.