Cyberbullying is an action related to the use of digital technology to intentionally hurt, humiliate or bully other people online. This research focuses on the classification of cyberbullying comments on social media, especially Instagram comments, where many parties who then become a group of people who don't like something will come together to provide negative opinions and comments, which can cause lowered self-confidence and other bad impacts for other users and account owner. Therefore, a classification of Instagram comments regarding cyberbullying was carried out as an effort to prevent this action. The data used in this research is 2000 data, where this data will go through various processes so that it can be executed. In this research, the Naïve Bayes method was used by dividing two classes, namely Bully and Not Bully. Based on the results of the tests that have been carried out, the results obtained are an accuracy value of 84%, a precision value of 84%, a recall of 84%, and an f1-score of 84%.