RECOGNISING CYBERBULLYING THROUGH MACHINE LEARNING TECHNIQUES: A COMPARATIVE ANALYSIS

Abstract

Millions of people use social media to communicate and share information with one another. Facebook, Twitter, and other social media platforms provide the ability to interact and converse with anybody, at any point in time, and with a huge group of individuals. On a global scale, social media is used by over three billion people. It is a 100% web-based platform that is constantly growing and evolving. According to the National Crime Prevention Council, cyberbullying occurs when individuals use their phones, online game programmes, or other electronic devices to email or send text, photographs, or videos with the intent of intentionally injuring or humiliating others (NCPC). Cyberbullying can occur at any time of day or week and can affect anyone who is online. Cyberbullying text messages, photographs, and videos can be anonymously posted and instantly distributed to a large audience. Maintaining track of those posts' reassessments may be difficult, if not impossible. Additionally, it is no longer possible to delete such communications after a specified time period has passed. Certain social networking sites offer a guide to avoiding cyberbullying. To protect your data, Facebook provides a section dedicated to detailing how to report cyberbullying and block the attacker. Users on Instagram can unfollow or block anyone who posts photographs or videos that make them feel uneasy. Violations of the Community Guidelines can also be reported directly from the app. According to Twitter, the consumer should be penalized for inappropriate, abusive, rude, or threatening behavior. Cyberbullying has been linked to social, emotional, and educational difficulties in adolescents, including despair and estrangement, as well as an increased risk of self-harm, including suicide. Numerous organizations are attempting to raise awareness about cyberbullying. In this thesis, we tested multiple machine learning models on text-based datasets.