Detection of Cyber bullying on Social Media Using Machine Learning

With the widespread use of the Internet in modern times, a massive amount of data has been generated. However, along with its advantages, the cyberworld also has its own set of disadvantages. One such disadvantage is cyberbullying, which is an online crime. Cyberbullying occurs when bullying takes place online using technology. This research paper presents review of30 different cyberbullying researchers and the methods they use to detect bullying. Cybercrime refers to all criminal activities that use the Internet as a means of access and are carried out through computers, mobile phones, and other electronic devices. The previous research on detecting cyberbullying has been limited due to factors such as unavailability of the dataset, hidden identities of the predators, and privacy of the victims. To overcome these limitations, an efficient text mining method is proposed in this paper, which uses machine learning algorithms to actively detect bullying texts. The dataset used for evaluation is collected from myspace.com and Preverted-Justice.com. Unlike previous studies that only considered textual features, this study extracted three types of features, namely textual, behavioral, and demographic features from the dataset. The text features include intimidating words that could lead to real cyberbullying results. The behavioral trait observed is that if a person is bullied once, they might bully someone else later. The demographic features extracted from the dataset include age, gender, and location.The system’s performance was evaluated using two classifiers, namely the Support Vector Machine (SVM) and the Bernoulli NB. The SVM classifier outperformed the Bernoulli NB with an overall accuracy of 87.1