Bangla Malicious Comment Detection in Social Media Posts Using Ensemble Learning

Information exchange and the growing usage of social media have greatly benefited human beings. However, social media often has a negative effect on freedom of speech and helps promote hatred, crimes, racial profiling, misogyny, and bullying. While reflecting public opinion and allowing for constructive criticism or admiration, the comment box on public websites can be seen by some as a platform to use obscene and abusive language without any consequences. It is considerably more difficult to identify malicious text in Bengali. In the age of technological automation, manually reviewing and eliminating offensive texts from publicly accessible websites may become tiresome. In this study, we used several machine learning algorithms and ensemble methods for identifying Bangla malicious text. We created a dataset by taking Bangla comments from Facebook and analyzed the effectiveness of different machine learning algorithms and ensemble methods in detecting malicious text accurately. According to the results of our tests, XGBoost, an ensemble technique, performed the best, with 70.23% accuracy in detecting malicious text.