Cyberbullying Detection and Identification Using Machine Learning-Based Hybrid Framework
Elham Alotaibi, Aida Al‐Samawi · IEEE Access · 2025
Cyberbullying has a significant impact on the many people who use social media, especially teenagers. Therefore, this study introduces a hybrid system that combines a support vector machine (SVM) and a long short-term memory (LSTM) network to detect cyberbullying on social networks, particularly on X (Twitter). The proposed model combines the deep contextual analysis capability of LSTM networks with the powerful classification performance of SVMs, using features extracted via the TF-IDF technique. A data set of English-language tweets was pre-processed and analyzed to train and consequently assess the hybrid system. The results demonstrate that the proposed approach achieves high accuracy and can effectively identify cyberbullying, even in subtle cases such as sarcasm. This study contributes to enhanced online safety by leveraging the strengths of both models in the hybrid framework, providing a scalable solution for real-time cyberbullying detection. Future recommendations include expanding the data set and exploring multi-modal approaches to improve the model’s detection capabilities.