Data-Driven Insights for Teaching Intervention: Profiling Student Types Through Clustering in a University Public Course

This study employs clustering algorithms to analyze multi-dimensional learning behavior data from a blended public course at Hanshan Normal University. Our analysis reveals a moderate positive correlation between process-oriented learning behaviors — such as video viewing, task completion, and check-in rates — and final exam scores. Through cluster analysis, we identify three distinct student profiles: Stable High-Achievers, To-Be-Improved, and Marginally-Stable, which exhibit significant differences in exam behavior, assignment submission patterns, and overall performance. We further determine that chapter task points, check-in rate, assignments, and chapter quizzes are the core factors influencing students’ comprehensive scores, whereas exams and course points contribute less. Notably, extreme assignment submission times coupled with very low scores (20-40 points) strongly associate procrastination and time management issues with academic underachievement. These findings offer data-driven insights for refining instructional design and implementing targeted student support.