Abstract
<jats:p>The purpose of this study was to assess the level of university teachers’ digital competence based on student evaluations and to cluster teachers with similar ratings using clustering methods. A total of 514 students from first- to fifth-year courses at seven universities (MNUMS, MUST, MNDU, MNUFE, Etugen, Ider, and Ach) participated in the survey, providing evaluations through seven questions. Clustering analyses were conducted using the k-means and k-medoids (PAM) methods. The results showed high reliability and suitability of the dataset for clustering, as evidenced by Cronbach’s alpha (≥0.947), the Silhouette coefficient (≥0.88), and the Hopkins statistic (≥0.965). Both methods optimally classified teachers’ digital competence into two groups, “low” and “high,” with average scores ranging from 2.2–2.6 and 3.9–4.4, respectively. Compared to k-means, the k-medoids method provided more robust results due to its medoid-based approach and resilience to outliers. Moreover, 36.2–39.1% of students’ responses fell into the “low” competence cluster, while 60.9–63.8% were categorized as “high,” highlighting the potential to differentiate teachers’ digital competence levels. One-way ANOVA further confirmed statistically significant differences between the two clusters across all seven items (F > 300, p < 0.001), validating the distinct evaluation patterns. Overall, the study demonstrates that clustering methods, based on student evaluations, provide a reliable data-driven approach for identifying the digital competence levels of university teachers.</jats:p>