Cyberbullying has become a serious social issue with the rapid growth of social media platforms. The massive volume of online user-generated content makes manual moderation ineffective and inconsistent. This research proposes an automated cyberbullying detection system using Machine Learning and Natural Language Processing techniques. Textual data is preprocessed and transformed using TF-IDF (Term Frequency–Inverse Document Frequency) vectorization to extract meaningful statistical features. Supervised classification algorithms such as Support Vector Machine (SVM) and Multinomial Naive Bayes are implemented to classify content as bullying or non-bullying.

Experimental evaluation demonstrates that TF-IDF combined with traditional machine learning classifiers provides an efficient and computationally lightweight solution for explicit abuse detection. The study also highlights the limitations of frequency-based approaches in handling sarcasm, contextual ambiguity, and evolving slang. The proposed model offers a scalable framework suitable for real-time deployment in social media platforms.

Keywords: Cyberbullying Detection, Machine Learning, TF-IDF, Natural Language Processing, Text Classification, SVM, Naive Bayes