Reinforcement Learning Interpretable Job Recommendation Based on Talent Knowledge Graph Reasoning

Abstract: In response to the difficulties of large-scale application, cold start, lack of novelty, and interpretability in existing job recommendation systems, a reinforcement learning interpretable recommendation method based on talent knowledge graph reasoning is proposed. Based on actual resume datasets, a talent social experience knowledge graph is constructed, and a policy agent is trained on the knowledge graph based on reinforcement learning theory. The recommendation process is decomposed into direction selection and node selection, enabling the agent to search for potential high-quality recommendation targets on the knowledge graph. Compared with LR, BPR.JRL-int, JRL-rep, and PGPR models, the reinforcement learning interpretable recommendation model based on talent knowledge graph reasoning performs best in MRR@20 (81.7%), Hit@1 (74.8%), Hit@5 (92.2%), and Hit@10 (97.0%). The limitations of the experimental dataset size and task type are relatively limited. The model effectively combines the historical work experience of talents and recommends based on similar work experience, while incorporating the correlation of job attributes in the knowledge graph. In addition to providing recommendation results, it also provides reasoning paths, which can effectively address the problems of cold start, lack of novelty, and interpretability. Keywords: job recommendation, knowledge graph reasoning, reinforcement learning interpretable recommendation.