An Ensemble Learning Framework for Robust Cyberbullying Detection on Social Media
Mohammed Hisham Saeed, Shakaib Ahmed Mohammed, Taufeeq Noamaan, Rania Mehreen Farooq, Mohammed Abdul Raheem · International Journal of Engineering and Advanced Technology · 2025
Social networking platforms on the Internet are now an essential feature of daily life worldwide, as these networks have made bridging the gap and sharing content an effortless task. Twitter stands out as a leading platform with a gigantic user base and is used extensively for communication between people and spreading information. Besides the many advantages these websites offer, such as promoting worldwide communication and dialogue, they may also pose unintended side effects that can be destructive to humanitarian and social life. One of the negative impacts of social networking sites is cyberbullying. Cyberbullying can be defined as “willful and repeated harm inflicted through the medium of electronic text” [1]. The support of harmful actions, such as harassment, threats, and humiliation, by individuals in online environments has brought about significant emotional and psychological effects for targeted individuals. The anonymity associated with social media platforms has the effect of increasing the occurrence of such detrimental activities, as there is less fear of the consequences of their actions, thus escalating the negative impact of cyberbullying. The Cyberbullying Detection Algorithm, a unique research approach, is used to combat the increasing problem of cyberbullying through ensemble-based learning algorithms, achieving a set of features for the Twitter dataset using machine learning techniques. This algorithm will look down on user-generated tweets in real time and discover patterns that may indicate cyberbullying behaviour. The role of the framework is to make the cyberbullying detection model on Internet platforms such as Twitter more accountable and effective through a mix of Machine Learning algorithms such as Random Forest, BERT, LSTM, and Ensemble. Our findings from an evaluative study of the critical features extracted from the Twitter dataset showed their relevance in cyberbullying detection. The performance evaluation based on key metrics such as F1 Score, Accuracy, AUC, and Precision depicts how the detection of cyberbullying can be made more effective and efficient by utilising machine learning algorithms that can detect online harassment and create a secure digital space for everyone.