한국지반공학회논문집 (Journal of the Korean Geotechnical Society)

  • 제40권5호
  • /
  • Pages.113-129
  • /
  • 2024
  • /
  • 1229-2427(pISSN)
  • /
  • 2288-646X(eISSN)
한국지반공학회 (Korean Geotechnical Society)
권호 표지
DOI QR코드

DOI QR Code

차원축소와 클러스터링 분석을 활용한 도로비탈면 위험등급 산정

Assessment of Risk Levels in Cut-Slope Using Dimensionality Reduction and Clustering Analysis

  • 서승환 (한국건설기술연구원 지반연구본부) ;
  • 김건웅 (한국건설기술연구원 지반연구본부) ;
  • 우용훈 (한국건설기술연구원 지반연구본부) ;
  • 박병석 (한국건설기술연구원 지반연구본부) ;
  • 김주형 (한국건설기술연구원 지반연구본부) ;
  • 김승현 (한국건설기술연구원 지반연구본부) ;
  • 정문경 (한국건설기술연구원 지반연구본부)
  • Seo, Seunghwan (Korea Institute of Civil Engr. and Building Tech.) ;
  • Kim, Gunwoong (Korea Institute of Civil Engr. and Building Tech.) ;
  • Woo, Younghoon (Korea Institute of Civil Engr. and Building Tech.) ;
  • Park, Byungsuk (Korea Institute of Civil Engr. and Building Tech.) ;
  • Kim, Juhyong (Korea Institute of Civil Engr. and Building Tech.) ;
  • Kim, Seung-Hyun (Korea Institute of Civil Engr. and Building Tech.) ;
  • Chung, Moonkyung (Korea Institute of Civil Engr. and Building Tech.)
  • 투고 : 2024.09.30
  • 심사 : 2024.10.10
  • 발행 : 2024.10.31
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

본 연구는 도로 비탈면 유지관리 데이터를 활용하여 위험등급을 재분류하고, 기존 위험등급 평가 방식의 한계를 분석하였다. 기존의 위험등급 평가는 주로 전문가의 주관적 판단에 의존하여 평가의 일관성에 한계가 있었다. 이를 개선하기 위해 본 연구에서는 차원 축소 기법인 주성분 분석(PCA)과 선형 판별 분석(LDA)을 적용하여 데이터를 축소하고, K-means 클러스터링을 통해 새로운 위험등급을 분류하였다. PCA를 사용한 클러스터링이 LDA에 비해 더 명확한 클러스터 분리를 보였으며, 실루엣 계수 등 성능 지표에서도 우수한 결과를 나타냈다. 이는 기존 위험등급 레이블이 데이터의 실제 구조를 충분히 반영하지 못함을 시사한다. 또한, LDA를 기반으로 한 클러스터링 결과와 기존 위험등급 레이블 간의 일치도가 낮아, 기존 레이블의 신뢰성에 한계가 있음을 확인하였다. 이를 해결하기 위해, PCA와 K-means 클러스터링을 이용하여 새로운 위험등급을 부여하고, 위험점수를 기반으로 각 클러스터의 등급을 분류하였다. 주요 위험 요인에 대한 정량적 분석을 통해 등급 부여의 기준을 설정하고, 각 등급별 주요 변수의 영향을 평가하였다. 본 연구는 데이터 중심의 객관적이고 정량적인 위험등급 평가 방법을 제시하여 도로 비탈면 관리의 효율성과 신뢰성을 향상시키는 데 기여하고자 한다.

This study reclassifies the risk levels of cut-slopes and addresses the limitations inherent in existing evaluation methods using road slope maintenance data. Conventional risk assessment predominantly relies on subjective expert judgment, resulting in issues of consistency and reliability. To mitigate these limitations, this study applies dimensionality reduction techniques, specifically Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA), followed by K-means clustering, to classify new risk levels. The clustering results using PCA demonstrated more distinct cluster separation compared to LDA, and also showed superior performance in terms of the silhouette coefficient and other clustering metrics. This suggests that the existing risk level labels may not adequately capture the underlying data structure. Furthermore, the inconsistency observed between LDA-based clustering results and current risk labels indicates potential reliability issues in the present labeling approach. To resolve this, new risk levels were assigned using PCA and K-means clustering, with cluster risk levels evaluated based on risk scores. A quantitative analysis of key risk factors was also conducted to establish criteria for risk classification and assess the impact of each variable on the different risk levels. This study proposes a data-driven, objective, and quantitative approach to risk level evaluation, aiming to improve the efficiency and reliability of road slope management.

키워드

과제정보

본 연구는 과학기술정보통신부 한국건설기술연구원 주요사업으로 수행되었습니다(과제번호 20240133-001, 지반분야 재난재해 대응과 미래 건설산업 신성장을 위한 지반 기술 연구).

참고문헌

  1. Alam, S. and Kwon, G.R. (2017), "Alzheimer Disease Classification Using KPCA, LDA, and Multi-kernel Learning SVM", International Journal of Imaging System and Technology, Vol.27, No.2, pp. 133-143.
  2. Ayesha, S., Hanif, M.K., and Talib, R. (2020), "Overview and Comparative Study of Dimensionality Reduction Techniques for High Dimensional Data", Information Fusion, Vol.59, pp.44-58
  3. Chung, H.S. and You, B.O. (1996), "Failure Types in Rock Slopes According to Geological Characteristics", Journal of the Korean Geotechnical Society, Vol.12, No.6, pp.37-50.
  4. Davies, D.L., and Bouldin, D.W. (1979), "A Cluster Separation Measure", IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.PAMI-1, No.2, pp.224-227.
  5. Dunn, J.C. (1974), "Well-Separated Clusters and Optimal Fuzzy Partitions", Journal of Cybernetics, Vol.4, No.1, pp.95-104.
  6. Erisoglu, M., Erisoglu, u., and Sakallioglu, S. (2013), "Effect of Dimension Reduction by Principal Component Analysis on Clustering", Journal of Statistics and Management Systems, Vol.14, No.2, pp.277-287.
  7. Ganesh, A., Maggs, B.M., and Panigrahi, D. (2023), "Universal Algorithms for Clustering Problems", ACM Transactions on Algorithms, Vol.19, No.2, pp.1-46.
  8. Gu, X., Nie, W., Geng, J., Yuan, C., Zhu, T., and Zheng, S. (2023), "Road Slope Monitoring and Early Warning System Integrating Numerical Simulation and Image Recognition: A Case Study of Nanping, Fujian, China", Stochastic Environmental Research and Risk Assessment, Vol.37, pp.3819-3835.
  9. Hermawan, F., Maulana, M., Ismiyati, Indarto, H., and Pamungkas, G. (2023), "Slope Risk Management System on Toll Road: A case study Semarang-Salatiga Section", The 8th International Conference on Energy, Environment, Epidemiology and Information System, Semaring City, Indonesia, Vol.448, No.01012, pp.1-9.
  10. Ikotun, A.M., Ezugwu, A.E., Abualigah, L., Abuhaija, B., and Heming, J. (2023), "K-means Clustering Algorithms: A Comprehensive Review, Variants Analysis, and Advances in the Era of Big Data", Information Science, Vol.662, pp.178-210.
  11. Ivosev, G., Burton, L., and Bonner, R. (2008), "Dimensionality Reduction and Visualization in Principal Component Analysis", Analytical Chemistry, Vol.80, No.13, pp.4933-4933.
  12. Jung, D.Y., Lee, S.M., Wang, H.M., Kim, J.H., and Lee, S.H. (2010a), "Fault Detection Method with PCA and LDA and its Application to Induction Motor", Journal of Central South University of Technology, Vol.17, pp.1238-1242.
  13. Jung, S.J., Kim, Y.S., and Kim, T.H. (2010b), "Risk Evlauation of Slope Using Principal Component Analysis (PCA)", Journal of the Korean Geotechnical Society, Vol.26, No.10, pp.69-79.
  14. Kim, S.H., Kim, H.G., Oak, Y.S., Lee, J.H., and Koo, H.B. (2013), "Analysis of Priority Investments for Preventing Roadside Slope Failures", Journal of Engineering Geology, Vol.23, No.3, pp.257-269.
  15. KICT (2021), Operation of Road Cut Slope Management System in 2021, Republic of Korea Ministry of Land, Infrastructure and Transport, pp.15-86.
  16. Lee, S.H., Kim, S.H., Woo, Y., Moon, J.P., and Yang, I. (2021), "The Study for Improvement of Data-Quality of Cut-Slope Management System Using Machine Learning", Journal of Engineering Geology, Vol.31, No.1, pp.31-42.
  17. Maulik, U. and Bandyopadhyay, S. (2002), "Performance Evaluation of Some Clustering Algorithms and Validity Indices", IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.24, No.12, pp.1650-1654.
  18. Nindya G.I., Abdul G., Ichsan, R., and Muhammad, T.Y.S. (2024), "GIS-Based Risk Assessment and Mitigation of Road Slope Instability in Morotai Island Regency", South Asian Research Journal of Engineering and Technology, Vol.6, No.1, pp.20-28.
  19. Ogbuabor, G. and Ugwoke, F. (2018), "Clustering Algorithm for a Healthcare Dataset Using Silhouette Score Value", International Journal of Cumputer Science and Information Technology, Vol.10, No.2, pp.27-37.
  20. Shin, C.G., Lee, S.Y., and Lee, S. (2008), "A Study on the Domestic Cut Slope Management Plan Reflecting Quality According to Rock Types", Journal of the Korean Geotechnical Society, Vol.24, No.2, pp.37-44.
  21. Varmuza, K. and Filzmoser, P. (2016), Introduction to Multivariate Statistical Analysis in Chemometrics, CRC Press, Boca Raton, pp.44-52.
  22. Vitalis, F., Muncan, J., Anantawittayanon, S., Kovacs, Z., and Tsenkova, R. (2023), "Aquaphotomics Monitoring of Lettuce Freshness during Cold Storage", Foods, Vol.12, No.2, 258.
  23. Woo, Y., Kim, S.H., Yang, I., and Lee, S.H. (2020a), "The Study for Utilizing Data of Cut-Slope Management System by Using Logistic Regression", Journal of Engineering Geology, Vol.30, No.4, pp.649-661.
  24. Woo, Y., Kim, S.H., Kim, J., and Park, G.H. (2020b), "Research on the Production of Risk Maps on Cut Slope Using Weather Information and Adaboost Model", Journal of Engineering Geology, Vol.30, No.4, pp.663-671.