All Issue

2026 Vol.44, Issue 3 Preview Page

Article

Predictive Model for Extinction Risk Areas with Machine Learning

머신러닝 기법을 이용한 소멸위험지역 예측모델 구축

Jungyeon KIM1
,
Myungsik DO2*
김 정연1
,
도 명식2*
1Master’s Degree, Department in Urban Engineering, Hanbat National University, Daejeon 34158, Korea
2Professor, Department in Urban Engineering, Hanbat National University, Daejeon 34158, Korea
1국립한밭대학교 도시공학과 석사연구원
2국립한밭대학교 도시공학과 교수
*Corresponding Author
30 June 2026. pp. 350-362
PDF XML Citation
Abstract

This study aims to build a machine learning-based predictive model to identify regions in South Korea at risk of local extinction. A Random Forest algorithm was applied to data from 232 cities, counties, and districts across the country. The dependent variable, the local extinction risk index, was calculated by averaging three existing indices related to regional population decline. To ensure a comprehensive analysis, 17 independent variables were selected from six domains: population, economy and finance, housing and households, transportation, environment and health care. The dataset was divided into training and testing subsets, and hyperparameters were optimized Using 5-fold cross-validation. The model showed strong predictive performance, with an R² of 0.94 and a MAPE of 7.76%, indicating reliable generalizability. Variable importance analysis revealed that the percentage of elderly population, OD traffic volume, and GRDP were the most influential factors in predicting extinction risk. Furthermore, time-series prediction results from 2025 to 2050 showed a gradual expansion of extinction risk, particularly among small and mid-sized cities outside the capital region. By 2050, over half of the regions were predicted to fall into high-risk categories. The consistency between actual and predicted values, with differences generally within ±1 risk stage, further supported the model’s external validity. These findings suggest that the proposed model can serve as a valuable tool for supporting long-term regional policy planning. However, a limitation of this study is that only six variables were available for future projections, excluding other potentially influential factors such as social or institutional elements. Future research should consider expanding the range of forecastable variables to enhance the model’s comprehensiveness and accuracy.

Keywords
extinction risk factors
extinction risk index
future projections
machine learning
random forest

본 연구는 한국의 지방소멸위험에 처한 지역을 식별하기 위한 머신러닝 기반 예측 모델을 구축하는 것을 목표로 하며, 전국 232개 시군구를 대상으로 랜덤포레스트(Random Forest) 알고리즘을 적용하였다. 종속변수인 지방소멸위험지수는 지역 인구 감소와 관련된 세 가지 기존 지수를 활용하였으며, 인구, 경제·재정, 주택·가구, 교통, 환경, 보건, 의료 여섯 가지 분야에서 총 17개 독립변수를 이용하였다. 데이터 셋은 훈련데이트와 테스트 데이트로 나누어졌으며 하이퍼파라미터는 5겹 교차검증을 통해 최적화하였다. 그 결과, 모델의 R2값은 0.94, MAPE 값은 7.76%로 모델의 일반화 능력이 신뢰할 수 있다. 변수별 중요도 분석 결과, 고령인구비율, OD통행량, GRDP가 소멸위험예측에 가장 큰 영향을 미치는 변수로 나타났다. 또한 2025년부터 2050년까지의 시계열 예측 결과, 특히 수도권 외 중소도시에서 소멸위험지역이 확산되는 경향을 보였으며, 2050년에는 절반 이상의 지역이 고위험지역으로 분류될 것으로 예측되었다. 본 연구의 성과는 장기적인 지역 정책 계획수립에 중요한 도구로 활용될 수 있을 것으로 기대된다. 그러나 장래예측 가능한 변수들이 일부에 국한되어 있으며, 사회적 또는 제도적 요인과 같은 다른 잠재적으로 영향력 있는 요소들을 충분히 반영하지 못한 점은 한계로 남아있다.

키워드
소멸위험요인
소멸위험지수
장래예측
머신러닝
랜덤포레스트
References
1

Breiman L. (2001), Random Forests, Machine Learning, 45(1), 5-32.

10.1023/A:1010933404324
2

Heo M. G., Lee S. H., Kim Y. S., Yoo I. S., Cho S. M., Kim J. Y., Cho S. G. (2022) Managing Population Decline in South Korea in an Era of Rural Decay Through Building a Virtuous Cycle of Regional Economy, Research Report, National Research Council for Economics, Humanities and Social Sciences, 22-44. (in Korean)

3

Hong J. U. (2021), A Mass Appraisal Model on Residential Property with Random Forest Algorithm, Journal of Real Estate Analysis, 7(1), 1-28.

10.30902/jrea.2021.7.1.1
4

James G., Witten D., Hastie T., Tibshirani R. (2013), An Introduction to Statistical Learning with Applications in R, New York: Springer.

10.1007/978-1-4614-7138-7
5

Ka S. J. (2023), Predictive Model for Resolving Public Conflicts Using Machine Learning Technique, The Korean Journal of Public Administration, 32(4), 225-251.

10.22897/KIPAJN.2023.32.4.008
6

Lee D. J. (2021), Factors Influencing Local Extinction: Focusing on 17 Metropolitan/Provincial Governments, Public Administration Journal, 33, 1-24. (in Korean)

7

Lee S. H. (2018), In-depth Analysis of Korea’s Local Extinction: Trends from 2013 to 2018 and Inter-regional Migration Outside the Capital Area, Employment Trend Brief, 2-21, Sejong: Korea Employment Information Service (KEIS). (in Korean)

8

Lim S. H., Hong S. J. (2019), Revitalizing Demand-Responding Mobility Service in Declining Local Cities, KOTI Basic Research Report, Sejong: Korea Transport Institute.

9

Loras-Gimeno D., Díaz-Lanchas J., Gómez-Bengoechea G. (2025), Rural Depopulation in the 21st Century: A Systematic Review of Policy Assessments, Regional Science Policy & Practice, 100176.

10.1016/j.rspp.2025.100176
10

Marcelino P., de Lurdes Antunes M., Fortunato E., Gomes M. C. (2019), Machine Learning Approach for Pavement Performance Prediction, International Journal of Pavement Engineering, 22(3), 341-354.

10.1080/10298436.2019.1609673
11

Masuda H. (2014), Local Extinction; Chuokoron-ShinSha, Tokyo.

12

Müller A. C., Guido S. (2016), Model Evaluation and Improvement, Introduction to Machine Learning with Python, 1, 262-263.

13

Pedregosa F., Varoquaux G., Gramfort A., Michel V., Thirion B., Grisel O. et. al. (2011), Scikit-learn: Machine learning in Python, Journal of Machine Learning Research, 12, 2825-2830.

14

Reades J., De Souza J., Hubbard P. (2019), Understanding Urban Gentrification Through Machine Learning, Urban Studies, 56(5), 922-942.

10.1177/0042098018789054
15

Statistics Korea (2024), 2024 Statistics on the Aged, Press Release, September 26, Retrieved from: https://kostat.go.kr/(Accessed: 26 September 2024).

16

Viñas C. D. (2019), Depopulation Processes in European Rural Areas: A case study of Cantabria (Spain), European Countryside, 11(3), 341-369.

10.2478/euco-2019-0021
17

Won G. H., Chae S. J., Seol Y. H. (2020), Is the Criterion for the Local Extinction Risk Index Appropriate?, Chungbuk FOCUS, 1-26. (in Korean)

18

Yoo H. B., Tak K. J., Mun J. S. (2021), Study on the Factors and Overcoming Methods of Extinction of Provinces in Korea: The Exploration with Machine Learning methods, The Korean Journal of Local Government Studies, 24(4), 443-476.

10.20484/klog.24.4.18
Information
  • Publisher :Korean Society of Transportation
  • Publisher(Ko) :대한교통학회
  • Journal Title :Journal of Korean Society of Transportation
  • Journal Title(Ko) :대한교통학회지
  • Volume : 44
  • No :3
  • Pages :350-362
  • Received Date : 2025-11-25
  • Revised Date : 2025-12-31
  • Accepted Date : 2026-02-20
  • DOI :https://doi.org/10.7470/jkst.2026.44.3.350
Journal Informaiton Journal of Korean Society of Transportation Journal of Korean Society of Transportation
Archives
: 2015 Vol.34~
  • scopus
  • NRF
  • KOFST
  • crossref cited-by
  • crosscheck
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close