딥러닝을 이용한 비명시적 괜찮은 일자리 예측과 이들의 지역적 특성 연구

초록

This study introduces the concept of implicit decent jobs and explores their regional characteristics. When quantitative indicators of job quality (e.g., wages, working hours, contract type, pension provision) are unavailable, a deep learning model is proposed to predict decent jobs using only job titles and descriptions. The model, based on XLM-RoBERTa and trained on 50,302 postings with explicit information from the WorkNet Open API, achieved 74.3% accuracy. It was then applied to 13,008 postings lacking such data, identifying 5,418 as implicit decent jobs. These jobs are more common in skill-intensive or flexible manufacturing sectors and tend to reflect the unique traits of their regions.

키워드