Real-Time Cyberbullying Detection Using NLP and Ensemble Learning Techniques

Cyberbullying, particularly on social media platforms, has become one of the serious menaces wherein the user, especially children and young adults, faces harassment and even emotional trauma through abusive or aggressive messages. Its detection is difficult because online language evolves, comprising sarcasm, slang, mixed languages like Hinglish, abbreviations, emoji, and between-the-lines meaning. The keyword-based traditional systems are unable to understand these complexities. The aim of this research work is to introduce an AI-based system for the detection of cyberbullying using Natural Language Processing (NLP) and supervised machine learning models that are deployed on real-time monitoring. This system preprocesses the social media text by removing noise and normalizing slang, thereby converting unstructured language into useful data. The resultant features are then classified as bullying or non-bullying based on the learned patterns. It presents the results from several classification models and shows that the Random Forest classifier performed best because of its strong generalization and reduction in prediction errors. The results prove the role that automated AI moderation can play in detecting such harmful content early, sending timely alerts, and thereby supporting a safer digital communication environment for all users.