Correlation Analysis Between Physical Activity and Psychological Assessment of College Students Based on Data Mining Techniques
Year: 2026
Type: conference-paper
Study design: Observational Study
Abstract: Cross-sectional screening and risk factor identification of college students‘ mental health issues are fundamental tasks in university mental health work. This study applies data mining techniques, combining cluster analysis, association rule mining, and supervised learning classification, to conduct a cross-sectional analysis of physical activity behavior and psychological assessment data from 1,006 Chinese college students. Cluster analysis uses a two-stage hybrid algorithm: first, HDBSCAN for noise point detection, followed by a Gaussian mixture model (GMM) fitted on non-noise samples, with noise points assigned to the nearest cluster based on posterior probability to classify the sample into four physical activity groups. Among these, the high-intensity regular exercise group shows significantly better comprehensive mental health scores than the low-activity high-sedentary group. Association rule mining using an optimized Apriori algorithm yields three significant rules: exercising at least four times per week is strongly associated with low depression risk (confidence 0.87); morning regular exercise combined with group exercise is moderately associated with high psychological resilience (confidence 0.73); and participation in group exercise is strongly associated with low social anxiety (confidence 0.82). Under a strict feature set (excluding psychological scale items), the random forest classifier achieves an AUC of 0.818 (95% CI: 0.792–0.843) for identifying severe psychological distress, and the XGBoost model achieves an AUC of 0.809. Common method bias testing shows that the first factor explains 23.7% of the total variance, within an acceptable range. This study provides a data-driven quantitative basis for cross-sectional mental health screening in universities and offers actionable indicators for designing personalized exercise intervention programs.
Author: Guoshuai Song
Institution: Dalian University of Finance and Economics
Language: English
Topic: Physical Activity and Health
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W7213245285