AUTHOR=Shi Xiuyu , Pan Jin , Yuan Daofu , Li Minye , Pan Yafeng 

TITLE=Advanced data analysis and prediction model for student mental health risk assessment

JOURNAL=Frontiers in Psychology

VOLUME=Volume 16 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2025.1682083

DOI=10.3389/fpsyg.2025.1682083

ISSN=1664-1078

ABSTRACT=With the increasing prevalence of mental health issues among students, early detection plays a crucial role in ensuring timely intervention. Existing methods struggle to capture the complex relationships among diverse data sources, such as behavioral, emotional, and physiological data, and fail to account for the temporal dynamics of mental health changes. This study addresses these challenges by proposing PsyGraph-SSL, a novel model that combines graph convolutional networks (GCN), temporal modeling, and self-supervised learning (SSL) to predict and analyze student mental health risks. The PsyGraph-SSL model integrates multi-modal data, including emotional, behavioral, and physiological signals, and learns temporal dependencies through time-series modeling. It employs GCN for processing social relationships and emotional interactions, while SSL is utilized to leverage unlabeled data and enhance feature learning. Temporal modeling further captures dynamic changes in students' mental health status, providing both short-term and long-term predictions. Experimental results on the WESAD and Student Well-Being Dataset show that PsyGraph-SSL outperforms traditional models, achieving higher accuracy, F1 score, AUC, and other key metrics. The model demonstrates strong performance in capturing emotional and behavioral fluctuations, making it highly effective for early detection and intervention. PsyGraph-SSL offers a comprehensive solution for student mental health monitoring, highlighting the importance of multi-modal data fusion and temporal analysis. The experimental results validate the model's potential for providing real-time, adaptive support. Future work will focus on expanding the dataset, improving generalization, and addressing challenges such as data imbalances and noise to further enhance the model's practical applicability.