Lecturers often recognize students at risk of receiving a grade E only after the final examination. This study explores whether Density-Based Spatial Clustering of Applications with Noise (DBSCAN) can identify academic inactivity patterns associated with grade E before the final examination. The data comprised 142 records of a machine learning course across two odd semesters of the 2024/2025 and 2025/2026 academic years, described by five pre-final features: participation, project, assignment, quiz, and midterm scores. In this dataset, every grade E coincided with a zero final-examination score, so "at risk" denotes non-completion of the final examination. DBSCAN parameters were selected using internal validity indices without using final grades. DBSCAN formed two clusters and 17 noise records. In a retrospective evaluation, the 25 records outside the main cluster included 21 of 28 grade-E records (precision 0.84, recall 0.75, F1-score 0.79, MCC 0.75). Isolation Forest performed identically, and cross-semester validation showed limited generalization (F1-score 0.39–0.40). DBSCAN therefore offers an interpretable exploratory approach for identifying inactivity patterns, not a validated early-warning system. Future work should use prospective validation and attendance or learning-log data.
Copyrights © 2026