Cyberbullying Detection on Social Networks using LSTM Model

Cyberbullying is a big problem in today’s society. The side effect of increasing social media, cyberbullying has emerged as a serious problem afflicting children, adolescents and young adults. Social networks provide a rich environment for bullies to uses these networks as vulnerable to attacks against victims. The development cyberbullying victims increases recently due to online harassment such as sharing private chats, rumours and sexual remarks take suitable actions to detect and prevent it. Several techniques are used to detect bullying messages in social media and its helps to construct secure social media surroundings. In previous Machine learning models they use various classification algorithms such as SVM, Navie Bayes, decision tree (DT) gives only an accuracy of 67.63%, 61.58%, 69.07% respectively. This research paper proposes Long Short Term Memory model (LSTM), a deep learning approach, is utilized for detecting and preventing cyberbulling actions. Extracting features using Long Short Term Memory, training the model and analyzing the model gives a more accurate result. The final evaluation of the proposed approach shows that LSTM achieves an accuracy of 75.12 % which is much better than previous models.