To address issues such as data sparsity, cold-start problems, and lack of interpretability in existing career recommendation methods, this paper proposes a career planning recommendation approach that integrates knowledge graphs with reinforcement learning.The method first constructs a unified knowledge graph that fuses multi-source information including users, positions, and skills.It then designs a hierarchical reinforcement learning inference mechanism that generates recommendations through multi-hop path exploration by agents within the graph, while simultaneously providing interpretable reasoning paths.Experiments on the public dataset career knowledge graph 15K demonstrate that our method achieves precision@10 of 0.782 and normalised discounted cumulative gain @10 of 0.815.Compared to the optimal baseline model, this represents significant improvements of approximately 6.3% and 5.1%, respectively.Notably, our approach exhibits enhanced robustness and interpretability, particularly in cold-start scenarios.
Meng Li (Thu,) studied this question.