DETECTING ONLINE HARASSMENT BASED ON SOCIAL MEDIA TEXT BY USING ENSEMBLE LEARNING

Abstract

Social media is now a tool for passing information, as well as a place to communicate but at the same time, a platform for cyberbullying, hate speech, and Offensive Language. In order to struggle this increasing problem, methods associated with machine learning, for example ensemble learning, are being employed to identify offensive material on such sites. A process with the name of Ensemble learning uses the predictions of many models of classifiers for instance Logistic Regression, SVC and Random Forest in order to classify and reduce the errors as much as possible. When researchers preprocess text from the social media, they were able to preprocess out such features as “Hate Speech,” “Cyberbullying,” and “Offensive Language.” With this approach, a study was accomplished up to 93% accuracy, meaning that ensemble learning is efficient at distinguishing online harassment and can be optimized more by other language model progressions plus sentiment analysis.