In today's era, amidst the rapid development of technology, many people misuse technology to do unpleasant things to others, including bullying that is done using social media called cyberbullying. Therefore, researchers classify social media comment data to determine whether it includes bullying or not. The purpose of this study is to classify social media comment data, including cyberbullying or not, by first comparing the performance between Naive Bayes Multinomial and Bernoulli algorithms in classifying such comment data. The researchers compared the Naive Bayes Classifier model, Multinomial and Bernoulli, to obtain the best model. The researchers also compared the use of the Bag of Words and TF-IDF feature extraction methods to improve the accuracy of the algorithms used. The results of the study show that the Naive Bayes Multinomial model algorithm obtained higher accuracy and faster average processing time compared to the Bernoulli model. The use of the Bag of Words feature extraction method can also significantly increase accuracy compared to TF-IDF.