Learning Job Competency Models from Historical Recruitment Data Using Supervised Machine Learning

Based on the structured representation of job responsibilities, qualification requirements, and skill information within historical recruitment texts, this study investigates data-driven learning methods for job competency models. The research focuses on feature representation of recruitment texts, construction of job competency labels, and training mechanisms for supervised learning. It details the processes of identifying competency elements, learning structural patterns, and constructing hierarchical structures. Experimental validation using a multi-position recruitment dataset yielded an accuracy of 0.851 on test samples and a macro-average F1 score of 0.832. Technical R&D positions achieved 0.892 accuracy, while operational management positions reached a macro-average recall of 0.842. Results demonstrate that the supervised learning framework reliably reconstructs job competency combinations.