A Comparative Analysis of Clustering Algorithms for Assessing College Students’ Key Competencies

The purpose of this study was to categorize students into distinct groups based on their key competency development levels. For this purpose, four clustering algorithms - Agglomerative hierarchical clustering, k-means clustering, Gaussian mixture model clustering, and DBSCAN - were utilized with data collected from 434 students at a university in Seoul. Agglomerative hierarchical clustering revealed two clusters, with one group showing higher development levels across all key competencies and the other exhibiting competencies below the average. Similarly, k-means clustering identified an optimal two-cluster solution. In contrast, Gaussian mixture model clustering and DBSCAN suggested a single cluster. Internal validity indices, such as the Davies-Bouldin index and the silhouette coefficient, favored k-means clustering with two clusters. Significantly, k-means clustering demonstrated higher stability compared to agglomerative hierarchical clustering. Consequently, this study selected k-means clustering with two clusters as the most suitable algorithm for the dataset, categorizing students into ‘Proficient’ and ‘Advanced’ groups.