Analysing Social Media Addiction and Deceptive Psychological Behaviour Among Students Using a Hybrid Deep Ensemble Framework

The growing incidence of social media addiction and manipulative psychological behaviours among students have caused worries over academic decline and mental health challenges. Conventional detection approaches encounter difficulties with contextual behavioural changes, requiring a mixed deep learning strategy. This study presents a Hybrid Deep Ensemble Framework (HDEF-SMBPA), which combines XGBoost for behavioural features with CNN for text-based sentiment and deception analysis. The system utilises SHAP-based explainable AI (XAI) to improve interpretability. Experimental findings indicate an accuracy of 94.2%, a recall of 93.5%, and a ROC-AUC of 0.97, exceeding those of conventional models. The findings underscore adverse academic associations, offering insights for psychologists and educators. The proposed system surpasses current approaches, providing a scalable, real-time solution for detecting and mitigating digital behavioural threats.