Experimental Evaluation of Robust Cyberbullying Detection over social media using Intelligent Learning Scheme

As the Internet has expanded, so too has the popularity of social media, which has become the preeminent means of communication in the modern day. While this improvement has helped people become more connected with one another, it has also led to some undesirable side effects. These include the emergence of harmful phenomena like cyberbullying, cybercrime, and online trolling. The negative impact of these issues cannot be underestimated, with instances of cyberbullying leading to severe mental and physical distress, particularly among vulnerable groups like women and children, sometimes even pushing them towards suicide. Similarly, the impact of online harassment has garnered significant attention due to its profound negative influence on social well-being. In this study, significant effort has been invested in reviewing prominent classification algorithms with a dual purpose: understanding their capabilities and limitations, and proposing an innovative approach to address the issue of cyberbullying. This approach employs Twitter datasets as a basis for detection. The proposed ensemble model constitutes a fusion of two distinct classifiers: the Decision Tree Classifier and the AdaBoost Classifier. The Proposed Model for cyberbullying identification demonstrates remarkable prowess. Its exceptional sensitivity score of 98.26% underscores its ability to accurately pinpoint positive cases, effectively identifying instances of cyberbullying.