All Issue

2026 Vol.42, Issue 1 Preview Page
XGBoost-Based Prediction of Sewer-Induced Sinkholes Under Imbalanced Data Conditions

불균형 데이터 처리를 고려한 XGBoost 기반 하수관 손상 유발 지반함몰 예측 모델

Joonyoung Kim1*
김 준영1*
1Member, Assistant Prof., Dept. of Artificial Intelligence, Hannam Univ.
1정회원, 한남대학교 AI융합학과 조교수
*Corresponding Author
28 February 2026. pp. 81-90
PDF XML Citation
Abstract

Prediction of ground subsidence caused by damaged sewer pipes is hindered by severe class imbalance due to the rarity of failure events. This study investigates the effect of imbalanced data handling strategies on machine-learning-based sinkhole prediction. An XGBoost classifier was adopted as the baseline model, and its performance without sampling was compared with that of models trained using random undersampling, random oversampling, and SMOTE under various class ratios. Model performance was primarily evaluated using PR-AUC, together with ROC-AUC, precision, recall, and F1-score. The results reveal that the baseline model achieved a relatively high ROC-AUC but relatively low PR-AUC, indicating limited capability in detecting rare sinkhole events. In contrast, random oversampling substantially improved minority-class prediction performance, with the highest PR-AUC observed at a class ratio of 1:5. These findings highlight the importance of PR-AUC-based evaluation and demonstrate that an appropriately designed oversampling strategy can effectively improve sinkhole prediction under highly imbalanced geotechnical data conditions.

Keywords
Sinkhole
Sewer
XGBoost
Imbalanced Data
Prediction

하수관 손상으로 인한 지반함몰 발생 예측은 발생 사례가 드문 특성으로 인해 극단적인 클래스 불균형 문제를 수반한다. 본 연구에서는 하수관 손상에 따른 지반함몰 발생 예측을 대상으로, 불균형 데이터 처리 기법이 기계학습 기반 예측 성능에 미치는 영향을 분석하였다. XGBoost 모델을 기본 분류기로 설정하고, 샘플링을 적용하지 않은 모델을 기준선으로 하여 random undersampling, random oversampling, SMOTE를 다양한 클래스 비율 조건에서 적용한 후 성능을 비교하였다. 성능 평가는 PR-AUC를 핵심 지표로 설정하고, ROC-AUC, 정밀도, 재현율, F1-score를 함께 고려하였다. 분석 결과, 기본 모델은 ROC-AUC 기준으로는 비교적 높은 값을 보였으나 PR-AUC는 상대적으로 낮아 소수 클래스 예측에 한계가 있음을 확인하였다. 반면, random oversampling을 적용한 모델은 소수 클래스 예측 성능이 유의미하게 향상되었으며, 특히 클래스 비율을 1:5로 설정한 경우 PR-AUC가 최대값을 나타냈다. 본 연구는 불균형 특성이 강한 지반함몰 예측 문제에서 PR-AUC 기반 평가의 중요성과 함께, 적절한 oversampling 전략의 효과를 실증적으로 제시한다.

References
1

Cardarelli, E., Cercato, M., De Donno, G., and Di Filippo, G. (2013), “Detection and Imaging of Piping Sinkholes by Integrated Geophysical Methods”, Near Surface Geophysics, Vol.12, No.3, pp.439-450, https://doi.org/10.3997/1873-0604.2013051.

10.3997/1873-0604.2013051
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), “XGBoost: A Scalable Tree Boosting System”, 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

Chen, W., Yang, K., Yu, Z., Shi, Y., and Chen, C. P. (2024), “A Survey on Imbalanced Learning: Latest Research, Applications and Future Directions”, Artificial Intelligence Review, Vol.57, No.6, 137, https://doi.org/10.1007/s10462-024-10759-6.

10.1007/s10462-024-10759-6
5

Friedman, J. H. (2001), “Greedy Function Approximation: A Gradient Boosting Machine”, Annals of Statistics, Vol.29, No.5, pp.1189-1232.

10.1214/aos/1013203451
6

Guo, S., Shao, Y., Zhang, T., Zhu, D. Z., and Zhang, Y. (2013), “Physical Modeling on Sand Erosion around Defective Sewer Pipes under the Influence of Groundwater”, Journal of Hydraulic Engineering, Vol.139, No.12, pp.1247-1257, https://doi.org/10.1061/(ASCE)HY.1943-7900.0000785.

10.1061/(ASCE)HY.1943-7900.0000785
7

He, H. and Garcia, E. A. (2009), “Learning from Imbalanced Data”, IEEE Transactions on Knowledge and Data Engineering, Vol.21, No.9, pp.1263-1284, https://doi.org/10.1109/TKDE.2008.239.

10.1109/TKDE.2008.239
8

Indiketiya, S., Jegatheesan, P., and Pathmanathan, R. (2017), “Evaluation of Defective Sewer Pipe Induced Internal Erosion and Associated Ground Deformation Using Laboratory Model Test”, Canadian Geotechnical Journal, Vol.54, pp.1184-1195, https://doi.org/10.1139/cgj-2016-0558.

10.1139/cgj-2016-0558
9

Kim, K., Kim, J., Kwak, T. Y., and Chung, C. K. (2018), “Logistic Regression Model for Sinkhole Susceptibility due to Damaged Sewer Pipes”, Natural Hazards, Vol.93, No.2, pp.765-785, https://doi.org/10.1007/s11069-018-3323-y.

10.1007/s11069-018-3323-y
10

Krawczyk, B. (2016), “Learning from Imbalanced Data: Open Challenges and Future Directions”, Progress in Artificial Intelligence, Vol.5, No.4, pp.221-232, https://doi.org/10.1007/s13748-016-0094-0.

10.1007/s13748-016-0094-0
11

Kuwano, R., Horii, T., Yamauchi, K., and Kohashi, H. (2010), “Formation of Subsurface Cavity and Loosening due to Defected Sewer Pipes”, Japanese Geotechnical Journal, Vol.5, pp.349-361.

10.3208/jgs.5.349
12

Min, D. H., Kim, Y., Kim, S., and Yoon, H. (2023), “Strategy of Oversampling Geotechnical Parameters through Geostatistical, SMOTE, and CTGAN Methods for Assessing Susceptibility of Landslide”, Landslides, Vol.21, No.12, pp.1-17, https://doi.org/10.1007/s10346-023-02166-9.

10.1007/s10346-023-02166-9
13

Park, J. H., Kim, J. B., Lee, S., Kang, J., and Mun, D. (2024), “Hybrid MLP-CNN-based Ground Sink Susceptibility Prediction in Urban Area Using Underground Pipe Map”, Reliability Engineering & System Safety, 245, 110031, https://doi.org/10.1016/j.ress.2024.110031.

10.1016/j.ress.2024.110031
14

Rogers, C. D. F. (1986), “The Mechanics of Internal Erosion”, Ground Engineering, Vol.19, No.3, pp.32-37.

15

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
16

Song, Y., Yang, D., Wu, W., Zhang, X., Zhou, J., Tian, Z., Wang, C., and Song, Y. (2023), “Evaluating Landslide Susceptibility Using Sampling Methodology and Multiple Machine Learning Models”, ISPRS International Journal of Geo-Information, Vol.12, No.5, 197, https://doi.org/10.3390/ijgi12050197.

10.3390/ijgi12050197
17

Velarde, G., Gómez, D., and Riquelme, J. C. (2024), “Tree Boosting Methods for Balanced and Imbalanced Classification and their Robustness Over Time in Risk Assessment”, Intelligent Systems with Applications, Vol.22, 200354, https://doi.org/10.1016/j.iswa.2024.200354.

10.1016/j.iswa.2024.200354
18

Yoon, H. (2023), “Method for Assessing Landslide Susceptibility Using SMOTE and Classification Algorithm”, Journal of the Korean Geotechnical Society, Vol.39, No.6, pp.5-12. (in Korean)

10.7843/KGS.2023.39.6.5
Information
  • Publisher :The Korean Geotechnical Society
  • Publisher(Ko) :한국지반공학회
  • Journal Title :Journal of the Korean Geotechnical Society
  • Journal Title(Ko) :한국지반공학회 논문집
  • Volume : 42
  • No :1
  • Pages :81-90
  • Received Date : 2026-01-09
  • Revised Date : 2026-01-26
  • Accepted Date : 2026-01-27
  • DOI :https://doi.org/10.7843/kgs.2026.42.1.81
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