Behavior-Transformer for Early Risk Detection and Monitoring in Students’ Social Media Activity

Abstract

In the digitalized world, examining students’ behavior through social media activity has become an important issue for organizations trying to detect early signs of depression, anxiety, and performance deterioration. Despite numerous records of inspirational guidance and rewards in educational environments, a significant gap remains in flexible, understandable strategies for real-time risk identification on professional social media platforms. We propose a Behavior-Transformer (Behav-T), a hybrid deep learning model designed to identify student mental health risk from self-reported social media behaviour.  Digital transformation activities necessitate and create opportunities for ethical monitoring of individuals. The results obtained by the proposed model show 0.77 accuracy and 0.77 F1-score, outperforming traditional models, including Support Vector Machine (SVM), Gradient Boosting, Logistic Regression, Random Forest, and a Multi-Layer Perceptron (MLP). The results show that the proposed model performs well for early mental health risk screening in student populations. Future work should use multimodal data and cross-cultural validation across organizational contexts to achieve equitable, culturally responsive risk detection.