A Review on Cyberbullying Detection Using Machine Learning

Abstract

This reveals yet another grave problem: cyberbullying, which is a variety of harmful repeated digital aggression using anonymity and very sophisticated AI machine learning techniques. Cyberbullying is indeed a very common malaise that digital platforms come across and its impact is huge on the minds and emotions of its victims. Our study is in the direction of building models and evaluating the machine for cyberbullying detection in text-based data.We have used various machine learning algorithms with which a Random Forest classifier led to an average of 94% accuracy. This was done through very aggressive data scrapping from the social media platform, followed by rigorous preprocessing, class balancing, and very stable model performance. As part of the study, ethical considerations lie in user privacy protection, where false positives were reduced to the maximum possible extent. Machine learning can effectively identify cyber bullying, creating a safer digital environment. Future research should focus on distance analysis, multimedia data expansion, slang recognition, and unsupervised learning techniques, with implications for technical innovations and societal ethics.