Research on Mental Health Portrait and Dynamic Early Warning System of University Students with Digital Intelligence Empowerment

Abstract

This study focuses on the cross-integration of digital intelligence technology and college students' mental health work, and puts forward a set of mental health portrait and dynamic early warning system for college students empowered by digital intelligence. At the level of portrait construction, we study the integration of multi-source heterogeneous data such as academic, behavioral, social, psychological evaluation and active reporting by mobile terminals, adopt the federal learning framework to realize data fusion under privacy protection, construct a dynamic psychological portrait from four dimensions: basic risk, behavioral dynamics, stress event detection and protection factors, and realize student group classification through K-means clustering. At the level of early warning system, a four-level dynamic early warning index system of red, orange, yellow and blue is designed, which integrates rule engine, isolated forest anomaly detection and LSTM-Attention time series prediction model to upgrade from static threshold to dynamic, multi-level and interpretable early warning. At the same time, a closed-loop response process is constructed to ensure timely follow-up after early warning. Based on the experimental data of 8,420 students in a university in one academic year, it shows that the coverage rate of the system is 21.2%, and the overall effective rate of early warning is 53.6%, among which the effective rate of red warning is 100%, which significantly improves the identification and intervention efficiency of psychological crisis. The research provides a theoretical framework and practical path for the digital transformation of mental health services in universities empowered by digital technology.