<![CDATA[Career readiness assessment increasingly requires competency-based profiling beyond academic achievement, yet conventional clustering methods remain vulnerable to outliers that distort student classification. This study proposes an outlier-aware hybrid DBSCAN–K-Means framework to map the career readiness of computer science students using eight non-academic competency dimensions. Data were collected from 566 students across 21 Indonesian universities using a validated 35-item questionnaire covering leadership, collaboration, time management, self-directed learning, goal setting, adaptability, problem-solving, and technical skills. Cluster quality was evaluated using the Elbow Method, Silhouette Score, and Davies–Bouldin Index. DBSCAN identified 113 outliers (19.96%), and removing these observations improved clustering performance, increasing the Silhouette Score from 0.292 to 0.326 while reducing the Davies–Bouldin Index from 1.229 to 1.122. The hybrid approach identified four meaningful career readiness profiles, including highly prepared students, students requiring competency development, critically underprepared outliers, and exceptionally high-performing outliers overlooked by conventional clustering. These findings demonstrate that outlier-aware clustering produces more robust competency profiles and provides a replicable analytical framework for evidence-based career development strategies in higher education.]]>
Copyrights © 2026
