Large-Scale Behavioral Network Analysis: Unveiling the Impact of the COVID-19 Pandemic on University Student Interactions

This study examines the structural evolution and behavioral regularity of student social networks before, during, and after COVID-19. Using 65 million campus smart-card transactions (2019--2024), we constructed temporal friendship networks and applied motif detection with behavioral orderliness analysis. Results show stage-specific impacts of the pandemic. During Spring 2020, clustering coefficients, entropy, and average degree dropped sharply, reflecting a collapse of established social structures. Within several months, networks largely rebounded, with 4-clique motifs (M3) indicating reliance on tight cliques for support. Under normalized restrictions (2021--2022), a hybrid structure of ``high clustering--long path'' emerged, combining offline recovery with persistent digital ties. Post-pandemic (2023--2024), networks became more fragmented, though M3 motifs persisted, suggesting durable reorganization. Behavioral orderliness followed a parallel trajectory: rapid decline in 2020, partial recovery in 2021--2022, and convergence at higher levels after reopening. Students whithin high-frequency dyads displayed greater orderliness, underscoring a potential bidirectional relationship between stable ties and routine. These findings highlight the fragility, adaptability, and reconstructive capacity of student social systems, offering data-driven evidence for designing resilience strategies in higher education under future crises.