HYBRID AI MODEL FOR CYBERBULLYING DETECTION IN SOCIAL MEDIA: A TWITTER CASE STUDY

Abstract

Cyberbullying has become a significant concern on social media platforms, particularly on Twitter, where users frequently engage in discussions that may contain harmful content. Traditional rule-based and machine learning approaches for detecting cyberbullying often struggle with high false positive rates and limited contextual understanding. This study proposes a Hybrid AI Model for Cyberbullying Detection in Social Media, integrating deep learning and natural language processing (NLP) techniques to enhance detection accuracy.
The model combines Convolutional Neural Networks (CNNs) and Bidirectional Long Short-Term Memory (BiLSTM) networks to extract both spatial and sequential patterns from tweets. Additionally, transformer-based models such as BERT (Bidirectional Encoder Representations from Transformers) are incorporated to improve contextual understanding. Feature engineering techniques, including sentiment analysis, lexical embeddings, and abusive word detection, further enhance classification performance.
Experimental results on publicly available Twitter datasets demonstrate that the proposed hybrid model outperforms traditional deep learning approaches in terms of accuracy, precision, and recall. The framework provides a scalable and efficient solution for real-time cyberbullying detection, contributing to a safer online environment. Future work will focus on real-time deployment, explainable AI integration, and cross-platform adaptability to enhance cyberbullying mitigation strategies.