Analysing Machine Learning Techniques for Cyberbullying Detection: A Review Study

Social media networks have turned into an essential piece of people's lives. The issue of cyberbullying has developed along with the growth of consumers on social media platforms. In this research, we conducted a thorough analysis of machine learning techniques for behavior analysis. These studies used supervised and unsupervised techniques to identify cyberbullying characteristics by matching text-based information with distinguished characteristics. In our comprehensive literature review, the approaches used by different studies can be categorized into four domains, namely supervised, unsupervised, and weakly supervised machine learning models. The paper also summarized the type of features and their combinations used to detect cyberbullying behaviors. We also discovered vital challenges faced when conducting cyberbullying research, such as mapping the definition of cyberbullying to traditional bullying, the lack of benchmark data sets, and the selection of content and user-based features. This study provides a significant map for future research to better direct their efforts.