An Deep Convolutional Neural Networks are used to Detect Cyberbullying on Social Networks.

The ramifications of cyberbullying, a horrific kind of inappropriate conduct that may take place online, are upsetting. Speech is the most common form of expression in social networks, despite the fact that it may take other forms. The automated identification of such instances requires the use of intelligent system components. The majority of the existing research has adopted a conventional approach to machine learning to address this problem; however, the majority of the models that have been constructed are exclusively relevant to a certain social network. Deep learning-based models are effective for the identification of cyberbullying events, according to recent study. These models claim to be able to overcome the limitations of standard models and improve detection performance. In this investigation, we focus on the most recent findings that have been published on the subject of interest. We were able to duplicate and validate the findings of this body of research by using the same datasets (Wikipedia, Twitter, and Formspring) as the authors of the original study. Next, we looked at how well the models worked on other social media platforms and applied the previously developed algorithms to a new dataset from YouTube, which consisted of 54,000 postings contributed by 4,000 different users. We also evaluated the performance of the trained models after moving them from one platform to another and then analyzed the results. Deep learning-based models from our company perform much better than machine learning models that were trained on the same data from YouTube. We believe it is essential to take into account the potential impact that individuals’ social media accounts might have on deep learning-based models and to include the relevant data into the overall picture.