Journal of Engineering, Project, and Production Management, 2026, 16(3), 2026-198
Psychological Abnormal Student Identification Based on Multi-Source Heterogeneous Educational Data
1 Associate
Professor, International School of Technical Education, Sichuan
University of Architectural Technology, Deyang, 618000, China.
2 Associate Professor,
International School of Technical Education, Sichuan University of
Architectural Technology, Deyang, 618000, China, E-mail:
huangjiaohj1@163.com (corresponding author).
Project Management
Received February 7, 2026; revised March 16, 2026; accepted March 19, 2026
Available online April 8, 2026
Abstract: With the increasingly prominent mental health issues among college students, identifying students with psychological abnormalities through behavioral data has become an urgent research problem to be solved. A Hybrid Model for Psychological Abnormal Student Behavior Identification (HMPABI) based on Multi-source Heterogeneous Educational Data is proposed. By combining multidimensional behavioral characteristics and psychological assessment data from students during their school years, an efficient and accurate identification model for students with psychological abnormalities is constructed through clustering, oversampling techniques, and a mixed classification strategy combining logistic regression and support vector machines. Psychological abnormalities are defined operationally as significant behavioral deviations in social, academic, and daily routines that correlate with established clinical indicators of mental health risks. The study takes the Urumqi University Student Campus Behavior Dataset and the Adolescent Mental Health and Behavior Dataset for experimental verification. Performance was evaluated using the Mean Absolute Error framework to quantify the deviation between predicted risk scores and actual assessment labels across different behavioral observation time windows, which represent the data aggregation intervals for anomaly detection. On the Urban Underground Space-Central Business District UUS-CBD dataset, the HMPABI model consistently had lower error than the comparison models across all testing windows, achieving a maximum error of 0.28. In contrast, the errors of the other two models reached 0.45 and 0.42 at the maximum time window (45 minutes), respectively. The HMPBAI model can fully explore potential information of students behavioral characteristics. By integrating different types of data, it can more accurately predict students with psychological abnormalities. This study provides a new technological path for mental health monitoring, wherein the dynamically generated risk scores can be integrated into campus support systems to actively alert counseling centers, thereby enabling targeted, proactive early interventions for at-risk college students.
Keywords: Psychological abnormalities, behavioral characteristics, K-means clustering, smote oversampling, machine learning, decision support systems, educational management, risk monitoring, data-driven intervention.
Copyright © Journal of Engineering, Project, and Production Management (EPPM-Journal).
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
Requests for reprints and permissions at eppm.journal@gmail.com.
Citation: Dong, H. and Huang, J. (2026). Psychological Abnormal Student Identification Based on Multi-Source Heterogeneous Educational Data. Journal of Engineering, Project, and Production Management, 16(3), 2026-198.
