А HYBRID COMPETENCY CLUSTERING METHODOLOGY FOR CURRICULUM STRUCTURE FORMATION

Abstract

Traditional curriculum development practice is characterized predominantly by em-pirical determination of disciplines and subjective workload distribution, which impedes the systematic integration of stakeholder requirements and the rational use of educational re-sources. The absence of formalized methodologies for transforming competencies into the-matically coherent groups for discipline formation impedes the systematic design of curric-ula aligned with labor market requirements.

This study develops a hybrid competency clustering methodology that systematical-ly transforms stakeholder expert assessments into structured educational clusters through quantitative methods. The methodology integrates weighted consensus analysis of employ-er, graduate, and faculty surveys with competency prioritization based on the Knowledge, Skills, and Abilities (KSA) model, and hybrid clustering combining Ward's hierarchical clustering, K-means optimization, and silhouette analysis for quality validation. A multi-component proximity matrix integrates the semantic similarity (TF-IDF), KSA profiles, and prerequisite relationships with cross-validation-determined weights.

Experimental validation on the Software Engineering undergraduate program (15 competencies, 290 respondents) demonstrates an optimal cluster count of k = 5 with a mean silhouette coefficient of 0.71±0.08 and high thematic coherence (0.680.82). Stability analysis confirms the robustness to parameter variations (Rand index >0.90) with a mini-mum sample size of n≥30 per group. The resulting clusters serve as the foundation for sub-sequent discipline formation and workload distribution.