Enhancing Social Media Safety with Machine Learning-Based Cyberbullying Detection

Social media platforms have revolutionized global communication but also facilitated the rise of cyberbullying, posing serious threats to user well-being, particularly among youth. Manual moderation is inadequate for managing the scale and velocity of harmful content online. This paper proposes a machine learning-based system for real-time cyberbullying detection, leveraging TF-IDF vectorization and a Logistic Regression classifier to identify and categorize user comments as toxic, obscene, threatening, or hateful. A Flask-powered web interface enables users to evaluate comment toxicity interactively. Designed for scalability and future expansion, the system supports integration with social media APIs, multi-language processing, and potential adoption of deep learning models. This work aims to contribute to safer online environments through intelligent, automated moderation.