Deep Learning based Methods for Cyberbullying Detection on Social Media

The rise of social media sparked a revolution in the fields of communication and digitalization. These platforms were developed with the goal of bringing together people from all over the world while enabling them to express themselves and absorb the thoughts of others. Additionally, given the variety of origins, beliefs, races, and cultures represented among site users, many of them have a tendency to use derogatory language and bully people. Automated cyberbullying detection is not only challenging, but it is also a pressing requirement, considering how essential social networks have become in people’s life and the serious consequences of cyberbullying, particularly among youth. To address this problem, we are identifying cyberbullying in the twitter dataset based on age, religion, gender, and ethnicity in this work. We have employed machine learning models Naive Bayes, Logistic Regression, SVM and ensemble machine learning models Random Forest and XGBoost while we have leveraged LSTM and GRU deep learning models as the dataset contains approximately 48k tweets. GRU outperformed all the other models with F-1 score 0.92.