Comparative analysis on deep neural network models for detection of cyberbullying on Social Media

Abstract

Social media usage has been increased and it consists of both positive and negative effects. By considering the misusage of social media platforms by various cyberbullying methods like stalking, harassment there should be preventive methods to control these and to avoid mental stress. These extra words will expand the size of the vocabulary and influence the performance of the algorithm. Therefore, we come up with variant deep learning models like LSTM, BI-LSTM, RNN, BI-RNN, GRU, BI-GRU to detect cyberbullying in social media. These models are applied on Twitter, public comments data and performance were observed for these models and obtained improved accuracy of 90.4%.