All Issue

2025 Vol.41, Issue 6 Preview Page
Evaluation of AI-Based Prediction for Ground Subsidence and Its Feature Importance Analysis Using Imbalanced Multivariate Data

불균형 다변량 데이터를 활용한 인공지능 기반 지반함몰 예측 성능 분석 및 변수 중요도 해석

YuJin Kim1
,
Hyobum Lee2
,
Sung Jin Lee3
,
SeonHong Na4*
김 유진1
,
이 효범2
,
이 성진3
,
나 선홍4*
1Member, Graduate Student, Dept. of Smart City Engineering, Inha Univ.
2Member, Senior Researcher, AI Transformation (AX) Division for Railroad & Transportation, Korea Railroad Research Institute
3Member, Principle Researcher, Track & Civil Infrastructure Division, Korea Railroad Research Institute
4Member, Associate Prof., Dept. of Civil Engineering, Inha Univ.
1정회원, 인하대학교 스마트시티공학과 석사과정
2정회원, 한국철도기술연구원 철도교통AX본부 선임연구원
3정회원, 한국철도기술연구원 궤도토목본부 책임연구원
4정회원, 인하대학교 사회인프라공학과 부교수
*Corresponding Author
31 December 2025. pp. 179-193
PDF XML Citation
Abstract

This study evaluates the probability of ground subsidence and examines the applicability of AI-based prediction models using publicly available data. Gwangju Metropolitan City was selected as the study area, and a spatial dataset was constructed using five years of ground subsidence records (2019–2023), rainfall data (including 7-day and 30-day cumulative rainfall), geological information, and the distribution of aged buildings. To address severe class imbalance, techniques including undersampling, the synthetic minority oversampling technique, class weighting, and focal loss were applied. A comparative analysis was then performed using Random Forest, XGBoost, and deep neural networks. Although overall F1-scores were constrained by the intrinsic rarity of subsidence events, precision–recall analysis demonstrated that the models achieved meaningful classification performance. Among the evaluated models, XGBoost exhibited the most stable and accurate predictive capability. Feature importance analysis identified cumulative rainfall and the number of aged buildings as the most influential variables, indicating that urban ground subsidence is predominantly governed by the combined effects of long-term rainfall accumulation and aging underground infrastructure.

Keywords
Artificial Intelligence
Data imbalanced
Ground subsidence
SHAP
XGBoost

본 연구에서는 공공데이터를 활용하여 지반함몰 발생 가능성 분석하고, 인공지능 기반 예측 모델의 적용성을 평가하였다. 연구 대상 지역으로 광주광역시를 선정하고, 최근 5년간(2019-2023) 보고된 지반함몰 사례를 기반으로 강수량, 노후건물수, 지질정보 등의 변수를 수집하여 공간 데이터셋을 구축하였다. 지반함몰 발생 비율이 극히 낮은 불균형한 데이터 특성을 고려하여 언더샘플링(Undersampling), SMOTE, 클래스 가중치(Class weight), Focal loss 등의 다양한 불균형 처리 기법을 적용하였으며, DNN, Random Forest, XGBoost 모델을 이용해 예측 성능을 비교·분석하였다. 분석 결과, 전체적인 F1-score는 낮게 나타났으나, 정밀도-재현율(Precision-Recall) 곡선을 통해 분류 성능이 유의미한 수준임을 확인하였으며, 트리 기반의 XGBoost 모델이 가장 안정적이고 우수한 성능을 보였다. 변수 중요도 분석에서는 누적 강수량(7일, 30일)과 노후건물수가 주요 영향인자로 도출되었다. 이는 지반함몰이 단기적 요인보다는 장기적인 누적 강우와 노후화된 인공 구조물의 복합적 상호작용에 의해 발생함을 시사한다.

References
1

Breiman, L. (2001), “Random Forests”, Machine Learning, Vol.45, No.1, pp.5-32.

10.1023/A:1010933404324
2

Chawla, N. V., Bowyer, K. W., Hall, L. O., and Kegelmeyer, W. P. (2002), “SMOTE: Synthetic Minority Over-sampling Technique”, Journal of Artificial Intelligence Research, Vol.16, pp.321-357, https://doi.org/10.1613/jair.953.

10.1613/jair.953
3

Chen, T. and Guestrin, C. (2016, August), “Xgboost: A Scalable Tree Boosting System”, In Proceedings of the 22nd acm Sigkdd International Conference on Knowledge Discovery and Data Mining (pp.785-794), https://doi.org/10.1145/2939672.2939785.

10.1145/2939672.2939785
4

Choi, C. H., Kim, J. Y., Kang, J. M., Lee, S. Y., and Baek, W. J. (2022), “Ground Subsidence Risk Analysis on Correlation between Rainfall and Rainfall Intensity”, Journal of The Korean Society of Agricultural Engineers, Vol.64, No.3, pp.75-83, https://doi.org/10.5389/KSAE.2022.64.3.075.

10.5389/KSAE.2022.64.3.075
5

Cho, Hong-Lae and Jeong, Jong-Chul (2006), “Application of Spatial Interpolation to Rainfall Data”, Spatial Information Research, Vol.14, No.1, pp.29-41.

6

Davis, J. and Goadrich, M. (2006, June), “The Relationship between Precision-Recall and ROC Curves”, In Proceedings of the 23rd International Conference on Machine Learning (pp.233-240), https://doi.org/10.1145/1143844.1143874.

10.1145/1143844.1143874
7

Kim, J., Kang, J., and Choi, C. (2021), “Correlation Analysis of the Occurrence of Ground Subsidence According to the Density of Underground Pipelines”, Journal of the Korean GEO-environmental Society, Vol.22, No.11, pp.23-29, https://doi.org/10.14481/jkges.2021.22.11.23.

10.14481/jkges.2021.22.11.23
8

Kim, J., Kang, J. M., and Baek, S. H. (2021), “Development of Random Forest Model for Sewer-induced Sinkhole Susceptibility”, Journal of the Korean Geotechnical Society, Vol.37, No.12, pp.117-125, https://doi.org/10.19077/tsoc.2021.38.05.

10.19077/tsoc.2021.38.05
9

Kim, K. (2021), “Noise Avoidance SMOTE in Ensemble Learning for Imbalanced Data”, IEEE Access, Vol.9, pp.143250-143265, https://doi.org/10.1109/ACCESS.2021.3120738.

10.1109/ACCESS.2021.3120738
10

Kim, Y., Hyo, B., Sung, J., and Na, S. (2025), “Ground Subsidence in South Korea and Limitations of its Management system”, Journal of the Korean Society of Hazard Mitigation, Vol.25, No.4, pp.129-136, https://doi.org/10.9798/KOSHAM.2025.25.4.129.

10.9798/KOSHAM.2025.25.4.129
11

Korea Ground Safety Information System (JIS). (2024), Ground Subsidence Statistics, Retrieved from https://www.jis.go.kr/

12

Korea Infrastructure Safety and Technology Corporation. (2015), Research on Development of Manual for Daily Signs and Safety Management Against Ground Subsidence.

13

Kumar, P., Bhatnagar, R., Gaur, K., and Bhatnagar, A. (2021, March), Classification of Imbalanced Data: Review of Methods and Applications, In IOP Conference Series: Materials Science and Engineering (Vol.1099, No.1, p.012077). IOP Publishing, https://doi.org/10.1088/1757-899X/1099/1/012077.

10.1088/1757-899X/1099/1/012077
14

Kwak, T. Y., Chung, C. K., Kim, J., Lee, M., and Woo, S. I. (2019), “Experimental Assessment for the Effect of Burial Depth on the Formation of Underground Cavities and Ground Cave-ins by Damaged Sewer Pipes”, Journal of the Korean Geotechnical Society, Vol.35, No.11, pp.37-49, https://doi.org/10.7843/kgs.2019.35.11.37.

10.7843/kgs.2019.35.11.37
15

Lee, S., Kang, J., and Kim, J. (2023), “Machine Learning-based Predictive Model of Ground Subsidence Risk Using Characteristics of Underground Pipelines in Urban Areas”, IEEE Access, Vol.11, pp.69326-69336, https://doi.org/10.1109/ACCESS.2023.3293089.

10.1109/ACCESS.2023.3293089
16

Lee, S., Kang, J., and Kim, J. (2024), “Prediction of Ground Subsidence Risk in Urban Centers Using Underground Characteristics Information”, Applied Sciences, Vol.14, No.23, 11044, https://doi.org/10.3390/app142311044.

10.3390/app142311044
17

Lin, T. Y., Goyal, P., Girshick, R., He, K., and Dollár, P. (2017), “Focal Loss for Dense Object Detection”, In Proceedings of the IEEE International Conference on Computer Vision (pp.2980-2988), https://doi.org/10.1109/ICCV.2017.324.

10.1109/ICCV.2017.324
18

Lundberg, S. M. and Lee, S. I. (2017), “A Unified Approach to Interpreting Model Predictions”, Advances in Neural Information Processing Systems, 30, https://doi.org/10.48550/arXiv.1705.07874.

10.48550/arXiv.1705.07874
19

Ministry of Land, Infrastructure and Transport. (2015), Safety Management Manual for Ground Subsidence (sinkhole).

20

Mulyanto, M., Faisal, M., Prakosa, S. W., and Leu, J. S. (2020), “Effectiveness of Focal Loss for Minority Classification in Network Intrusion Detection Systems”, Symmetry, 13(1), 4, https://doi.org/10.3390/sym13010004.

10.3390/sym13010004
21

QGIS Development Team. (2024), QGIS Geographic Information System (Version 3.34) [Computer Software]. Open Source Geospatial Foundation. https://www.qgis.org.

22

Richardson, E., Trevizani, R., Greenbaum, J. A., Carter, H., Nielsen, M., and Peters, B. (2024), “The Receiver Operating Characteristic Curve Accurately Assesses Imbalanced Datasets”, Patterns, Vol.5, No.6, https://doi.org/10.1016/j.patter.2024.100994.

10.1016/j.patter.2024.10099439005487PMC11240176
23

Ryu, D. W., Yum, B. W., Lee, H. J., Jeong, S. W., and Jung, B. (2017), “Real-scale Experimental Simulation to Identify a Mechanism for Urban Subsidence”, Journal of the Korean Society of Civil Engineers, Vol.65, No.9, pp.34-40.

24

Saito, T. and Rehmsmeier, M. (2015), “The Precision-recall Plot is More Informative than the ROC Plot when Evaluating Binary Classifiers on Imbalanced Datasets”, PloS one, Vol.10, No.3, e0118432, https://doi.org/10.1371/journal.pone.0118432.

10.1371/journal.pone.011843225738806PMC4349800
25

Seoul Metropolitan Government. (2023), Underground Safety Impact Assessment Service for the Seoul Metropolitan Subway Line 9 Phase 4 Extension Construction, Seoul Metropolitan Government Urban Infrastructure Headquarters.

26

Shrestha, N. (2020), “Detecting Multicollinearity in Regression Analysis”, American Journal of Applied Mathematics and Statistics, Vol.8, No.2, pp.39-42, https://doi.org/10.12691/ajams-8-2-1.

10.12691/ajams-8-2-1
Information
  • Publisher :The Korean Geotechnical Society
  • Publisher(Ko) :한국지반공학회
  • Journal Title :Journal of the Korean Geotechnical Society
  • Journal Title(Ko) :한국지반공학회 논문집
  • Volume : 41
  • No :6
  • Pages :179-193
  • Received Date : 2025-11-17
  • Revised Date : 2025-12-24
  • Accepted Date : 2025-12-24
  • DOI :https://doi.org/10.7843/kgs.2025.41.6.179
Journal Informaiton Journal of the Korean Geotechnical Society Journal of the Korean Geotechnical Society
Archives
: Vol.23~28 (2007~2012)
  • NRF
  • KOFST
  • KISTI Current Status
  • KISTI Cited-by
  • crosscheck
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close