Classification cyber harassment on Twitter using multinomial naïve bayes

The study uses the Multinomial Naïve Bayes (MNB) method to classify four types of cyber harassment on Twitter, namely Physical Threats, Purposeful Embarrassment, Racist, and Sexual Harassment. A total of 2,000 Indonesian language tweets were used as samples and have been manually labeled using training data and testing data. The classification results show that the model achieves an accuracy of 77%, with a consistent accuracy value of 76.21% based on the K-fold cross-validation test. This study shows that MNB is effective in multiclass text classification to detect cyber harassment and provides a computationally efficient solution to the real-time content moderation system.