An Efficient Method for Detecting Cyberbullying Using Supervised Machine Learning Techniques

• Published in: Procedia computer science Vol. 258; pp. 1254 - 1261
• Main Authors: Joshi, Bansidhar, Joshi, Bineet Kumar, Pant, Sangeeta, Kumar, Anuj, Sharma, Hitesh Kumar
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 2025
• Subjects: Cyberbullying; Machine Learning (ML); Natural Language Processing (NLP); Term Frequency-Inverse Data Frequency (TF-IDF); Term Frequency-Inverse Data Frequency (TF-IDF); Machine Learning (ML); Cyberbullying; Natural Language Processing (NLP)
• DOI: 10.1016/j.procs.2025.04.359
• ISSN: 1877-0509

The issue of cyberbullying is more worrisome on social media platforms. Individuals are taking advantage of the unrestricted ability to express themselves on social media platforms to engage in this undesirable conduct. Although there are methods available to address this issue, they are subject to restrictions and may not employ optimal techniques. This research work aims to develop novel approaches to detect cyberbullying incidents automatically in real-time across various social media platforms, including tweets, comments, and messages. Using real-time Twitter data, including headlines, comments, and SMS messages from trending posts, we developed a labelling framework for cyberbullying research. We then analyzed this labeled dataset to explore the relationships between various traits associated with cyberbullying and cyber aggression, employing supervised machine learning (ML) and natural language processing (NLP) techniques. It is identified that linear support vector Classification (SVC) and stochastic gradient descent (SGD) classification algorithm are the most effective in classifying and predicting bullying messages in English language. The proposed solution is effective and rational, and it could offer a substantial contribution to the problem of detecting cyberbullying.