Big Data Analysis of Student Mental Health Based on Artificial Intelligence and Feature Extraction

View at publisher Summarise with VIRAT
Authors: Lingling Ji, Tianran Li, Xiaomeng Wu
Cited by: 0
References: 12

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

As student mental health issues become increasingly prominent and traditional screening methods face bottlenecks such as time lag, limited coverage, and missed detection of concealed problems, there is an urgent need to build a dynamic and precise early identification and intervention system. This study proposes an improved AI-driven solution. First, multi-source heterogeneous data are integrated and correlated features are extracted. Second, a Heterogeneous Graph Neural Network (HGNN) model is constructed, dynamically modeling psychological risk propagation paths through a graph attention mechanism. Finally, the SHAP (Shapley Additive exPlanations) interpretability framework is introduced to generate personalized psychological profiles. Experimental results show that the multimodal model integrating text, behavioral, and physiological features achieves an identification accuracy of 89.2%. Furthermore, the graph neural network excels in identifying propagation paths. Interpretability analysis reveals that physiological modal features provide a reliable basis for precise interventions. This research provides technical support for building a precise campus psychological early warning system and advances the early identification and intervention of mental health risks.

Fields of Study & Subjects

Educational Technology and Pedagogy Mental Health via Writing Online Learning and Analytics

Publication details

Journal / source
IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
Pages
195–200
Publication date
28 January 2026
Language
EN