Digital Resilience: A Review of Cutting-Edge Approaches to Cyberbullying Detection in the Social Media Landscape

Cyberbullying, characterized as a form of harmful online conduct, involves an ongoing pattern where a bully sends a series of abusive messages to a victim with the intention of causing harm over a period. Impropriety involving blackmailing, threatening, harassment, abusing, impersonating, mocking, etc. has become a widespread problem on social media platforms. To address this persistent issue, it's vital to develop a mechanism that can accurately detect and prevent cyberbullying. In this study, we present a comparison of cutting-edge machine learning and deep learning models that are applicable in detecting bullying instances in social media networks. It also examines the current landscape of advanced techniques and algorithms capable of detecting obscene language. Various neural network architectures, natural language processing techniques, ensemble methods, and hybrid systems are considered. A strong emphasis on the feature engineering process along with feature selection algorithms to select a significant feature based on human-engineered observations are also discussed.