Detecting and Preventing Cyberbullying Among Higher Education Students on social media with XGBoost and Particle Swarm Optimization

Social media has rendered cyberbullying a significant concern for college students. In contrast to high school student research, college campus research is limited and ambiguous. Many school districts have enacted cyberbullying prohibitions, although university campuses lack similar legislation. This initiative seeks to govern cyberbullying in higher education and inform educators about technological hazards. Unstructured data for the research of cyberbullying on social media was handled by text encoding, metadata filtering, and the deletion of non-linguistic elements. The creation of a Student slang and the elimination of superfluous stop words diminished the corpus's dimensionality. Information Gain and Chi-Square testing were integrated to discern characteristics. This work introduces an XGAPSO model that employs adaptive PSO and hyperparameter tuning to identify cyberbullying. The model surpassed its competitors, with an average F1-score of 97.62%, a recall of 93.50%, a precision of 95.18%, and an accuracy of 97.52%. This research demonstrates that XGAPSO is capable of identifying instances of cyberbullying among college students on social media platforms. Future research should enhance the model's applicability and bolster existing initiatives to address collegiate cyberbullying.