Driven by digital technology, this study systematically constructs an innovative model for student management in colleges and universities, and proposes a full chain management methodology through demand analysis, functional modeling and dynamic behavior mining. Based on the actual needs of colleges and universities, a system architecture covering authority control, multi-role interaction and dynamic configuration capability is designed. The improved K-means clustering algorithm is used to classify student learning habits into clusters, and lagged sequence analysis (LSA) is used to reveal the temporal correlation of learning behaviors. The empirical analysis relies on the behavioral logs and questionnaire data of MOOC platform, and finds that the sequence of learning behaviors is significantly correlated with course grades, e.g., the correlation coefficient of participating in the assessment after reviewing is 0.244, and the correlation coefficient of negatively affecting the grades by overlearning the new content is -0.298. The characteristics of campus network use classify the students into academic focus type 384, recreational and social type 1,032, balanced multitasking type 1807, and light use type 1,000. 1807 and light-use type 169, and their traffic distribution showed significant differences. During the semester, there are key turning points in behavioral patterns (week 5 and week 12), and students’ behaviors show stage-bystage evolution, e.g., group C dynamically adjusts from “diligent mode” to “result-oriented”. Data-driven behavioral modeling and time-series analysis can provide a scientific basis for personalized resource recommendation, precise intervention and dynamic management strategy optimization, and help university student management transform to intelligence and refinement.
