Cyberbullying Detection Using CNN Prediction Model

Cyberbullying has emerged as a pressing issue in today's digital era, with negative effects on individuals and society. Traditional approaches to address the issue have proven insufficient, thus, necessitating the exploration of novel solutions. This study focuses on the potential of deep learning algorithms to identify and prevent instances of cyberbullying. By using advanced algorithms and analysing large-scale datasets, this study aims to uncover patterns that help in the identification of cyberbullying incidents. The research utilizes a dataset containing around 47,692 labelled tweets consisting of six classes namely Religion-related cyberbullying, Gender-related cyberbullying, Ethnicity-related cyberbullying, Age-related cyberbullying, Not-cyberbullying and other type of cyberbullying. The balanced representation of these classes allows for a more thorough examination of the problem, making our approach more effective in addressing the various forms of cyberbullying. In the literature review, it was found that Deep Learning techniques consistently outperformed traditional Machine Learning methods in cyberbullying detection. This paper builds upon that knowledge by making use of Deep Learning to provide even more precise results. A novel CNN model is developed for an accurate and efficient detection model. The findings of the work contribute to an efficient cyberbullying detector with an accuracy of 96.7% and an AUC of 94.8% which will assist in the development of proactive measures to combat the negative impacts of cyberbullying.