CDNN: A Novel Methodology Development to Detect Online Social Network Cyberbullying Threats Using Convoluted Deep Neural Network Principle

Advanced methods for effective detection and prevention are required due to the growing prevalence of cyberbullying on online social networks. The Convoluted Deep Neural Network Principle (CDNN) is employed in this research to introduce a novel method for identifying cyberbullying threats. In order to extract hierarchical features from a variety of data inputs, the CDNN framework incorporates multiple layers of convolutional networks. The proposed model is cross-validated with the conventional classifier called Support Vector Machine (SVM) to evaluate the efficiency of the proposed scheme. The model captures the complex structures typical of cyberbullying behavior by processing these inputs through a successive convolution and pooling layers. We have performed a series of in-depth experiments to test this model with the ability to identify or classify cyberbullying incidents using a dataset that contained all manner of social networking interactions. The results we get prove that the performance of the CDNN-based approach outperforms traditional methods and other related deep learning models, in which it shows the detection accuracy rate as high as 98.02% detected by our model. More importantly, the robustness of CDNN toward different types of contents provided for cyberbullying can be mainly enhanced by its ability to handle multimodal data sources. Results emphasize the effectiveness of CDNN in tackling very complex issues related to recognition of cyberbullying and the precise platform for monitoring online activities.