Improving Graduate Course Assessment of Student Performance with DBSCAN and Weighted Averaging: a Case Study

Recent graduate course assessment methods have faced criticism for their inability to comprehensively evaluate student performance, particularly in courses where active participation and practical knowledge are crucial. Traditional frameworks often overlook key aspects of student engagement and learning, leading to inaccurate assessments. This paper presents a novel approach utilizing DBSCAN clustering and weighted averaging to evaluate student performance more holistically in the graduate-level course Geotechnical Engineering Theory and Practice. The method integrates multiple data points, such as attendance, punctuality, participation in lectures (both theoretical and case-based), and the quality of coursework. All features are categorized into three performance groups: learning engagement, active learning, and knowledge comprehension. DBSCAN clustering identifies distinct student groups based on participation patterns, while weighted averaging refines the evaluation by adjusting the relevance of each data point. The results from case study demonstrate the effectiveness of this approach in improving the accuracy and granularity of student assessments, offering valuable insights into course quality and providing a potential model for future course evaluation methods.