Analisis Sentimen pada Steam Review Menggunakan Multinomial Naïve Bayes dan Seleksi Fitur Gini Index Text

Video game is one of the entertainment medias chosen by most people today, many of which are played through computer devices. On computer devices, many video games are obtained through one of the game distribution platforms, namely Steam. However, Steam has several shortcomings, including those related to Steam reviews. On Steam reviews, you can see the rating of the game, but the rating does not really show the actual quality or condition of the game. As one example, there are users who give a high rating to a game, but in the comments column the user actually mentions the shortcomings of the game. To reduce or anticipate unclear reviews for users who want to try or buy the game, sentiment analysis on reviews is used. In this research, the output produced is information on the results of sentiment classification in filtering reviews, using the Multinomial Naïve Bayes algorithm and combined with the Gini Index feature selection. Sentiment classification is divided into two classes, namely recommended and not recommended classes. In this study, to test the sentiment classification system, a dataset containing reviews in the form of review sentences from Steam is used. The test results using Multinomial Naïve Bayes and Gini Index, can achieve the best accuracy of 60.29%.