Annals of Occupational and Environmental Medicine

Original Article

Occupation classification model based on DistilKoBERT: using the 5th and 6th Korean Working Condition Surveys: Tae-Yeon Kim, Seong-Uk Baek, Myeong-Hun Lim, Byungyoon Yun, Domyung Paek, Kyung Ehi Zoh, Kanwoo Youn, Yun Keun Lee, Yangho Kim, Jungwon Kim, Eunsuk Choi, Mo-Yeol Kang, YoonHo Cho, Kyung-Eun Lee, Juho Sim, Juyeon Oh, Heejoo Park, Jian Lee, Jong-Uk Won, Yu-Min Lee, Jin-Ha Yoon; Ann Occup Environ Med 2024;36:e19. Published online August 6, 2024; DOI: https://doi.org/10.35371/aoem.2024.36.e19

Abstract Abstract in Korean PDF Supplementary Material PubReader ePub: Background
Accurate occupation classification is essential in various fields, including policy development and epidemiological studies. This study aims to develop an occupation classification model based on DistilKoBERT.
Methods
This study used data from the 5th and 6th Korean Working Conditions Surveys conducted in 2017 and 2020, respectively. A total of 99,665 survey participants, who were nationally representative of Korean workers, were included. We used natural language responses regarding their job responsibilities and occupational codes based on the Korean Standard Classification of Occupations (7th version, 3-digit codes). The dataset was randomly split into training and test datasets in a ratio of 7:3. The occupation classification model based on DistilKoBERT was fine-tuned using the training dataset, and the model was evaluated using the test dataset. The accuracy, precision, recall, and F1 score were calculated as evaluation metrics.
Results
The final model, which classified 28,996 survey participants in the test dataset into 142 occupational codes, exhibited an accuracy of 84.44%. For the evaluation metrics, the precision, recall, and F1 score of the model, calculated by weighting based on the sample size, were 0.83, 0.84, and 0.83, respectively. The model demonstrated high precision in the classification of service and sales workers yet exhibited low precision in the classification of managers. In addition, it displayed high precision in classifying occupations prominently represented in the training dataset.
Conclusions
This study developed an occupation classification system based on DistilKoBERT, which demonstrated reasonable performance. Despite further efforts to enhance the classification accuracy, this automated occupation classification model holds promise for advancing epidemiological studies in the fields of occupational safety and health.; DistilKOBERT를 기반으로 한 직업 분류 모델 개발: 제5차, 6차 한국근로실태조사를 이용하여
목적
정확한 직업분류는 정책 개발 및 역학 연구를 포함한 다양한 분야에서 중요하다. 본 연구는 자연어처리모델인 DistilKoBERT를 기반으로 한 직업 분류 모델을 개발하는 것을 목표로 한다.
방법
본 연구는 2017년과 2020년에 실시된 제5차와 제6차 근로환경조사 (KWCS)의 데이터를 활용하였다. 대한민국 근로자를 국가적으로 대표하는 총 99,665명의 참가자가 포함되었고, 직무 내용과 관련된 자연어 응답과 그에 맞는 대한민국 표준직업 분류코드(7차 개정, 3자리 코드)를 연구에 사용하였다. 데이터셋은 7:3의 비율로 훈련 및 테스트 데이터셋으로 무작위로 분할되었고, 사전 학습된 DistilKoBERT을 훈련 데이터셋을 통해 파인튜닝하여 모델을 학습시키고, 테스트 데이터셋을 사용하여 그 기능을 평가하였다. 정확도, 정밀도, 재현율 및 F1 점수가 평가 지표로 계산되었다.
결과
테스트 데이터셋의 28,996명의 참가자를 142개의 직업 코드로 분류한 최종 모델은 84.44%의 정확도를 보였다. 샘플 크기를 기준으로 가중치를 적용하여 계산한 모델의 정밀도, 재현율 및 F1 점수는 각각 0.83, 0.84 및 0.83 이었다. 최종 모델은 서비스, 판매 종사자 그룹에서 높은 정밀도를 보여주었지만 관리자 그룹에서는 낮은 정밀도를 보였다. 또한 훈련 데이터셋에서 표본의 수가 많았던 직업에서 대체로 높은 정밀도를 보였다.
결론
본 연구는 DistilKoBERT를 기반으로 합리적인 성능을 보이는 직업 분류 모델을 개발하였다. 분류의 정확성을 향상시키기 위한 추가적인 노력이 필요하지만, 자동화된 직업 분류 모델은 직업 안전 및 보건 분야의 유행병 연구를 발전시키는 데 기여할 것이라 기대된다.

221 View
18 Download

Data Profile

Data resource profile: the Korean Working Conditions Survey (KWCS): Yoonho Cho; Ann Occup Environ Med 2023;35:e49. Published online November 23, 2023; DOI: https://doi.org/10.35371/aoem.2023.35.e49

Abstract Abstract in Korean PDF PubReader ePub

The Korean Working Conditions Survey (KWCS) is a state-approved statistical survey that has been conducted by the Occupational Safety and Health Research Institute (OSHRI) every 3 years since 2006 to monitor changes in the working conditions of Koreans. This cross-sectional national survey involves a sample of 50,000 employed people aged 15 or older. KWCS measures various working conditions through > 130 survey questions, including questions regarding working hours, labor intensity, work–life balance, degree of exposure to risk factors, and subjective health status. Professional survey interviewers visit households and conduct face to face interviews. KWCS provides data and statistics for occupational safety and health polices and research in Korea. Furthermore, OSHRI holds academic conferences every year, awards high-quality academic papers, and supports researchers using data. Microdata is publicly available through the OSHRI website (https://oshri.kosha.or.kr).

자료원 프로파일: 근로환경조사
KWCS는 국가승인통계로 대한민국에 거주하는 만15세 이상 취업자 5만명을 표본으로 매 3년마다 실시하는 단면적 조사로, 2006년부터 산업안전보건연구원에서 실시하고 있다. KWCS는 전문조사원에 의해 가구를 방문하여 1:1 면접 조사로 이루어지며, 근무시간, 노동강도, 일과 삶의 균형, 유해위험요인에 대한 노출 정도, 주관적 건강상태 등 다양한 근로환경에 대해 130여개 이상의 문항에 대해 조사한다. 조사 결과는 정책 수립과 연구자를 위한 기초 자료로 제공하고 있으며, 마이크로데이터는 산업안전보건연구원 홈페이지(OSHRI.OR.KR)를 통해 공개하고 있다.

Citations

Citations to this article as recorded by

Association between multiple jobs and physical and psychological symptoms among the Korean working population
Seok-Yoon Son, Jin-Young Min, Seung-Woo Ryoo, Baek-Yong Choi, Kyoung-Bok Min
Annals of Occupational and Environmental Medicine.2024;[Epub] CrossRef
Commuting time, working time, and their link to insomnia symptoms among Korean employees: A cross-sectional study
Hoje Ryu, Suhwan Ju, Hye-Eun Lee, Seong-Sik Cho
Sleep Health.2024; 10(4): 434. CrossRef
Interaction between old age and precarious employment on poor mental well‐being in workers: Evidence from a nationally representative sample in Korea
Seong‐Uk Baek, Yu‐Min Lee, Jong‐Uk Won, Jin‐Ha Yoon
International Journal of Geriatric Psychiatry.2024;[Epub] CrossRef
Occupation classification model based on DistilKoBERT: using the 5th and 6th Korean Working Condition Surveys
Tae-Yeon Kim, Seong-Uk Baek, Myeong-Hun Lim, Byungyoon Yun, Domyung Paek, Kyung Ehi Zoh, Kanwoo Youn, Yun Keun Lee, Yangho Kim, Jungwon Kim, Eunsuk Choi, Mo-Yeol Kang, YoonHo Cho, Kyung-Eun Lee, Juho Sim, Juyeon Oh, Heejoo Park, Jian Lee, Jong-Uk Won, Yu-
Annals of Occupational and Environmental Medicine.2024;[Epub] CrossRef
Long working hours, work-life imbalance, and poor mental health: a cross-sectional mediation analysis based on the sixth Korean Working Conditions Survey, 2020–2021
Seong-Uk Baek, Yu-Min Lee, Jin-Ha Yoon, Jong-Uk Won
Journal of Epidemiology.2024;[Epub] CrossRef
Association between sickness presenteeism and depressive symptoms: a cross-sectional study using the 6th Korean working conditions survey
Myeong-Hun Lim, Min-Seok Kim, Seong-Uk Baek, Tae-Yeon Kim, Jong-Uk Won, Jin-Ha Yoon
International Archives of Occupational and Environmental Health.2024; 97(5): 537. CrossRef
The association of job training duration and risk of depression among wage workers: an analysis of the mediating factors
Dong Geon Kim, Dong Kyu Kim, Kiook Baek
Annals of Occupational and Environmental Medicine.2024;[Epub] CrossRef

114 View
1 Download
7 Crossref

First
Prev
Page of 1
Next
Last

Search