Comparison of Multinomial Naïve Bayes and Bernoulli Naïve Bayes on Sentiment Analysis of Kurikulum Merdeka with Query Expansion Ranking

Social media is one of the public services for conveying or obtaining news, opinions and comments on an issue. One of the social media that is in great demand by the people of Indonesia is Twitter. Kurikulum merdeka is one of the most discussed issues currently on Twitter. Kurikulum merdeka is a curriculum that incorporates varied intra-curricular learning with more optimal content to provide students adequate time to investigate ideas and build expertise. Until now, kurikulum merdeka still reaps the pros and cons. To process and analyze further regarding opinions on the kurikulum merdeka, it can be done using sentiment analysis. The high dimension of features in the classification process becoming a problem in sentiment analysis because it causes classification to be inefficient, so feature selection is needed to solve this problem. The purpose of this study was to obtain the results of the classification of kurikulum merdeka sentiments using the multinomial naïve bayes and bernoulli naïve Bayes, as well as query expansion rankings for feature selection and to compare the performance of the two classifications. Multinomial naïve bayes classification produces 106 tweets with positive sentiment and 164 tweets with negative sentiment with accuracy, recall, precision and f-measure respectively 98.889%, 98.131%, 99.057% and 98.591%, while bernoulli naïve bayes produces 95 tweets with positive sentiment and 175 tweets with negative sentiment with accuracy, recall, precision, and f-measure respectively 94.815%, 87.850%, 98.947% and 93.069% respectively. Therefore, multinomial naïve bayes classifies the kurikulum merdeka sentiment better than bernoulli naïve bayes.