An Evolutionary Multi-Objective Optimization Framework for Career Path Recommendation Using Knowledge Graphs

Abstract

Finding the right career path has become increasingly challenging in today’s rapidly evolving labor market. Most existing job recommender systems rely on keyword matching or collaborative filtering, which often fail to capture the deeper relationships between skills, occupations, and long-term career
objectives. Moreover, they typically optimize a single metric, ignoring the fact that real career choices involve multiple trade-offs such as immediate fit, learning effort, and future potential. In this paper, we propose KG-CPR, a knowledge-graph-based framework for multi-objective career path
recommendation. Our approach integrates a rich knowledge graph constructed from the O*NET database with an evolutionary optimization engine that evaluates career options based on three objectives: maximizing skill alignment, minimizing the learning gap, and maximizing job potential.
Instead of collapsing these criteria into one score, KG-CPR identifies Pareto-optimal sets of career paths, ensuring that users receive diverse and strategically meaningful recommendations. We evaluate KG-CPR using synthetic user profiles generated from O*NET data and compare different optimization
strategies, including NSGA-II, MOEA/D, and our proposed MOEA/D-TS algorithm. Results show that multi-objective optimization methods significantly outperform single-objective baselines in both accuracy and diversity metrics. While MOEA/D achieved slightly higher ranking quality, MOEA/D
TS provided the best overall balance, delivering superior precision, recall, and diversity. These findings highlight the potential of KG-CPR as a practical tool for career advising and strategic workforce planning