Application of a Heterogeneous Graph for Analyzing Workforce Competency Networks

Recruitment in outsourcing companies requires effective matching between candidate competencies and job requirements. This study constructs and analyzes a heterogeneous candidatecompetencyposition graph using real recruitment data from an outsourcing company to examine the structural roles of competencies and their community organization. The recruitment database comprised 28,289 records collected from October 2025 to August 2026, with 875 unique candidates across five selected positions: Fitter, Maintenance, Mechanical Technician, Electrical Technician, and Welder. Using stratified, criterion-based purposive sampling with equal allocation, 20 candidates were selected from each position, resulting in 100 candidates. The resulting graph contained 115 vertices and 665 edges, consisting of 634 candidatecompetency relationships and 31 competencyposition relationships. Normalized betweenness centrality ranked Technical Drawing Reading highest (CB = 0.197077), followed by Welding (0.185315) and Teamwork (0.141790). Louvain community detection with resolution 1.0, weight=None, and random seed 38 identified five communities with modularity Q = 0.079237. A 30-run stability analysis produced three to six communities, with a mean pairwise ARI of 0.137730, indicating variation across random initializations. The findings are interpreted as exploratory structural patterns rather than statistically representative community structures.