Enhancing Cyberbullying Detection: A Comprehensive Framework Using Machine Learning, Deep Learning, and NLP Techniques

Cyberbullying has emerged as a significant societal issue in the digital age, causing emotional and psychological harm to individuals on social media platforms. The anonymity and reach of these platforms amplify harmful behaviors, necessitating effective interventions. Traditional moderation techniques, such as manual review or user reports, are insufficient to handle the growing scale and complexity of online content. This study introduces a comprehensive framework for cyberbullying classification, leveraging advanced machine learning (ML) techniques and natural language processing (NLP). It highlights the innovative application of SMOTETS to mitigate class imbalance and improve model robustness. A variety of classification models, including ensemble methods and deep learning approaches, were evaluated, with the highest accuracy reaching 92%. Emphasis is placed on preprocessing strategies like tokenization, stemming, and word embeddings to capture nuanced linguistic patterns, such as sarcasm and slang, commonly used in harmful content. By demonstrating the effectiveness of ensemble learning techniques, this work contributes to the development of scalable solutions for cyberbullying detection. The findings aim to advance research in automated abuse detection and promote safer digital environments for all users.