Sentiment analysis with a unique dataset shaped text is positive, negative or neutral with expressing words in media social shall be careful and follow the insult or done can result in exposed for those who commit criminal law violations related to post on social media. A method of classification as naive bayes(nb) proposed by many researchers for use in the analysis sentiment text. Of the bayes naive, will be tested with three input that is using tokenize (remove) punctuation, the cases transform instead (change and a small type) stopwords (took the word negative) (100 of commentary is positive and negative comments text) (100) text commentary. The research using bayes naive obtained the auc: 0.484 + / - 0.107 micro (:0.484) (:class it with the positive) 72,50 accuracy , hence a method bayes can produce classifications naive enough to analysis. Keywords: Ethics post, Naïve Bayes, Social Media.