Exploring Machine Learning Algorithms for Sentiment Analysis in the Sphere of Cyberbullying Detection

The rise in cybercrimes targeting children is leading to devastating consequences like harmful behaviors and even suicide. While social media platforms offer reactive solutions like post deletion and blocking, a more proactive approach is needed. To combat cyber child abuse effectively we need technology-driven solutions that focus on detection and alerts to identify predators or protect victims. Various strategies have been proposed, including machine learning techniques that leverage sentiment analysis to detect patterns in the bully's language. This study explores the potential of machine learning to combat cyber child abuse by focusing on detection and alerts. We investigated the effectiveness of supervised, unsupervised, and ensemble learning algorithms for sentiment analysis in identifying abusive language. Analyzing a social media dataset for bullying detection, the research reveals that Support Vector Machines achieved the highest accuracy (83.26%) among supervised algorithms, with K-Nearest Neighbor (82.97%) and Random Forest Classifier (87.74%) performing best in the unsupervised and ensemble categories, respectively. The paper also addresses limitations and proposes future research directions.