Detection of Cyberbullying on Social Media Using Machine Learning

Abstract

Online harassment disturbance by cyberbullying has grave repercussions. The majority of social networks have it in textual format and it exists in various forms. It is undeniable that more than 1.96 billion people are affected by social interactions online. However, as we move into the next decade, there are significant challenges to be faced in regulating user behaviour. The prevalence of harassment, bullying, and victimization has become a serious problem that needs to be addressed. Smart systems are needed to automatically detect such incidents. Recent studies have been focused on this issue by using conventional machine learning models, and many of the models developed in these studies can be applied to a single social network at a time and are capable of scaling up. Deep learning-based models have been found to be effective in identifying digital harassment incidents, surpassing the limitations of traditional models and improving detection performance. Although there are several conventional models available to manage such incidents, accurately categorizing the type of harassment still remains a challenge. To effectively use machine learning and language preparation to stop the violent outcome and screen the harassment in the virtual area, a framework is suggested to provide two distinct descriptions of cyberbullying. Our technique utilizes an innovative CNN-based approach for content analysis. However, many existing methods use a simplistic approach that provides less accuracy in the classification process. A current dataset is used for testing, and when our approach is compared to other methods currently in use, it is discovered to have higher precision and grouping.