Course assessment data hold learning achievement patterns that final grades alone do not reveal. This study aims to cluster students’ learning achievement patterns in a Database Management Systems course using the K-Medoids algorithm. The data consist of the scores of 97 anonymized students on six assessment components (participation, project, assignment, quiz, midterm, and final exam), normalized with Min-Max scaling. K-Medoids based on Partitioning Around Medoids (PAM) was tested for k = 2 to 6 with Euclidean and Manhattan distances, compared with K-Means, and evaluated using the Silhouette Coefficient, the Davies-Bouldin Index (DBI), and the Calinski-Harabasz Index. The results show that k = 3 is the most suitable solution (Silhouette 0.690; DBI 0.499), yielding identical partitions under both distances and remaining stable across 100 subsample tests (ARI = 1.00). The three clusters are consistently high achievers (70 students), high achievers without submitted assignments (18 students), and inactive students at risk of failing (9 students). Because medoids are actual students, the cluster profiles are easy to interpret for learning evaluation and early intervention.
Copyrights © 2026