Campus restorative environmental perception and academic resilience: An explanatory-predictive study using PLS-SEM and machine learning

Academic resilience supports university students in sustaining learning and adapting to academic difficulties. Drawing on Attention Restoration Theory (ART) and Conservation of Resources (COR) theory, this study examines how campus restorative environmental perception relates to academic resilience through perceived social support, place dependence, and place identity. Cross-sectional survey data from 506 university students were analyzed using partial least squares structural equation modeling (PLS-SEM), complemented by six machine learning classifiers. After accounting for the mediators, compatibility retained a positive direct association with academic resilience (β = 0.101, p = 0.010), whereas the direct paths from being away, fascination, and extent were not statistically significant. All four restorative dimensions showed significant indirect associations through each of the three mediators. Place dependence had the largest estimated path coefficient for academic resilience among the resource-related constructs (β = 0.301, p < 0.001). In the supplementary median-split classification, logistic regression achieved a cross-validated accuracy of 72.13%, compared with a majority-class baseline of 57.91%; place dependence was the first splitting variable in the fitted decision tree. These findings link students' perceptions of restorative campus spaces to social, functional, and emotional resources relevant to academic resilience. They suggest that campus design should combine activity-space compatibility with dependable places for study and recovery, opportunities for supportive interaction, and meaningful campus experiences.