International Journal of New Developments in Education, 2026, 8(2); doi: 10.25236/IJNDE.2026.080206.
Shangfei Lu
School of Education, Guangxi Vocational Normal University, Nanning, 530007, Guangxi, China
As mental health issues in colleges and universities become increasingly complex and hidden, traditional methods of mental health monitoring that rely on questionnaires and manual interviews have significant shortcomings in terms of data coverage, dynamic identification capabilities, and privacy protection, making it difficult to achieve continuous perception and accurate early warning of college students' mental state. To this end, this paper proposes a multimodal hierarchical early warning system for the mental health of college students in Guangxi based on federated learning, and explores its collaborative application mechanism in ideological and political intervention. A multimodal psychological feature modeling method is constructed, integrating learning behavior data, campus life behavior data, online interaction data, and psychological assessment data. Through feature representation learning and a multimodal fusion model, psychological characteristics such as students' emotional fluctuations, behavioral abnormalities, and learning pressure are extracted. An intelligent matching mechanism for ideological and political resources based on student psychological profiles is constructed to realize the transformation of psychological early warning results into personalized ideological and political intervention strategies. Experimental results show that the multimodal fusion model achieves an accuracy of 0.913 and an F1-score of 0.894 in the mental health identification task, which is about 10%–14% higher than the single-behavioral data model. In terms of collaborative training performance, the federated learning model achieves an accuracy of 0.912, reduces the single-round communication volume to 7.3 MB, and shortens the total training time to 20.1 min, balancing model performance with data privacy protection requirements.
College Student Mental Health Early Warning; Ideological and Political Intervention in Universities; Multimodal Feature Modeling; Federated Learning; Intelligent Risk Classification Model
Shangfei Lu. Development of a Multimodal Hierarchical Early Warning System for College Students' Mental Health Based on Federated Learning and Its Application in Ideological and Political Intervention. International Journal of New Developments in Education (2026), Vol. 8, Issue 2: 39-46. https://doi.org/10.25236/IJNDE.2026.080206.
[1] Zhang Z. Early warning model of adolescent mental health based on big data and machine learning [J]. Soft computing, 2024, 28(1): 811-828.
[2] Perochon S, Di Martino J M, Carpenter K L H, et al. Early detection of autism using digital behavioral phenotyping[J]. Nature Medicine, 2023, 29(10): 2489-2497.
[3] Kirkbride J B, Anglin D M, Colman I, et al. The social determinants of mental health and disorder: evidence, prevention and recommendations[J]. World psychiatry, 2024, 23(1): 58-90.
[4] O'Connor R C, Worthman C M, Abanga M, et al. Gone Too Soon: priorities for action to prevent premature mortality associated with mental illness and mental distress[J]. The Lancet Psychiatry, 2023, 10(6): 452-464.
[5] Uhlhaas P J, Davey C G, Mehta U M, et al. Towards a youth mental health paradigm: a perspective and roadmap[J]. Molecular psychiatry, 2023, 28(8): 3171-3181.
[6] Garg M. Mental Health Analysis in Social Media Posts: A Survey: M. Garg[J]. Archives of Computational Methods in Engineering, 2023, 30(3): 1819-1842.