Get started
Work

Identifying Students in Need of Psychological Support: An AI-Driven Approach Using Public Behavioral Data

Year: 2025
Type: preprint
Abstract: Background : This study investigates the use of machine learning to identify university students potentially in need of psychological support, using behavioral data from academic records, hygiene compliance, and disciplinary history. These data were collected from student-led administrative departments responsible for monitoring academic discipline, hygiene practices, and behavioral compliance within the campus environment. LightGBM was benchmarked against traditional models and achieved superior performance (AUC = 0.972, F1 = 0.875), supporting its potential as an ecologically valid, scale-free screening tool. Two key findings emerged. First, we identified a department-specific optimal behavioral observation period, with model performance peaking between Weeks 7 and 15 depending on the type of behavioral data, suggesting that mental health-related signals require a sufficient accumulation period to become predictive. Second, longitudinal prediction revealed that individuals in the at-risk group showed dynamic, individualized changes in their predicted mental health states, highlighting the need for continuous and personalized monitoring. These findings contribute to the development of AI-based early warning systems for psychological services in educational settings. Methods : Behavioral and mental health data from 1,255 students at a Chinese university were collected with informed consent. Mental health assessments, based on the Chinese College Students’ Mental Health Scale, served as supervised labels, while public behavioral data, covering academic achievement, behavioral performance were used as independent variables. Three typical supervised AI learning models (Logistic Regression, K-Nearest Neighbors, and LightGBM) were tested for classification performance. Results : A dataset of students’ academic achievement, behavioral performance, and psychological health status was constructed. Among three models, Light- GBM demonstrated the best performance, achieving a Precision of 0.875, Recall of 0.875, F1-score of 0.875, Accuracy of 0.967, and an AUC value of 0.972,closely aligning with the results of the mental health scale. Compared to Logistic Regression and K-Nearest Neighbors, Light-GBM exhibited superior predictive performance, particularly in AUC and Accuracy. Conclusion : This study confirms that supervised machine learning can serve as a real-time alternative to traditional psychological assessment methods, enhancing the efficiency of university’s counseling service departments . Additionally, two key findings emerged: (1) determining the minimum data collection period required for effective assessment of mental health status, optimizing university mental health service resource management; and (2) there are significant dynamic change patterns in the psychological states of students who may need psychological support, providing crucial insights for more precise personalized interventions. Future research should explore the generalizability of these findings across different university settings and student populations.
DOI: 10.21203/rs.3.rs-6399521/v1
Language: English
Open Access status: green
OpenAlex ID: W4409308519
Cites: 1
Cited by:
Related to: 10