All Issue

2020 Vol.30, Issue 3 Preview Page

Research Article

Prediction of Landslides and Determination of Its Variable Importance Using AutoML

AutoML을 이용한 산사태 예측 및 변수 중요도 산정

KoungHoon Nam1
,
Man-Il Kim2
,
Oil Kwon3
,
Fawu Wang4
,
Gyo-Cheol Jeong5*
남 경훈1
,
김 만일2
,
권 오일3
,
왕 파우4
,
정 교철5*
1Post Dr. Researcher, Geoscience and Geoenvironmental Science, Shimane University, Japan
2Principal Researcher, Forest Engineering Research Institute, National Forestry Cooperative Federation
3Research Fellow, Road Infrastructure Project Team, Korea Institute of Civil Engineering and Building Technology
4Professor, Department of Civil Engineering, Tongji University, China
5Professor, Department of Earth and Environmental Sciences, Andong National University
1일본 시마네대학 지구과학 및 지구환경과학과 박사후 연구원
2산림조합중앙회 산림공학연구소 책임연구원
3한국건설기술연구원 연구위원
4중국 퉁지대학 토목공학과 교수
5안동대학교 지구환경과학과 교수
*Corresponding Author
30 September 2020. pp. 315-325
PDF XML Citation
Abstract

This study was performed to develop a model to predict landslides and determine the variable importance of landslides susceptibility factors based on the probabilistic prediction of landslides occurring on slopes along the road. Field survey data of 30,615 slopes from 2007 to 2020 in Korea were analyzed to develop a landslide prediction model. Of the total 131 variable factors, 17 topographic factors and 114 geological factors (including 89 bedrocks) were used to predict landslides. Automated machine learning (AutoML) was used to classify landslides and non-landslides. The verification results revealed that the best model, an extremely randomized tree (XRT) with excellent predictive performance, yielded 83.977% of prediction rates on test data. As a result of the analysis to determine the variable importance of the landslide susceptibility factors, it was composed of 10 topographic factors and 9 geological factors, which was presented as a percentage for each factor. This model was evaluated probabilistically and quantitatively for the likelihood of landslide occurrence by deriving the ranking of variable importance using only on-site survey data. It is considered that this model can provide a reliable basis for slope safety assessment through field surveys to decision-makers in the future.

Keywords
AutoML
XRT
landslide prediction model
probabilistic prediction
variable importance

이 연구는 도로 비탈면에서 발생하는 산사태의 확률론적 예측에 기반된 산사태 발생에 영향을 미치는 인자의 중요도 산정 및 예측 모델을 개발하는 것이다. 산사태 예측 모델을 개발하기 위해 한반도 전 지역을 대상으로 2007년부터 2020년까지 조사된 30,615사면의 현장조사 자료를 활용하였다. 전체 131개의 변수 인자 중 지형인자 17개, 지질인자 114개(기반암 89개를 포함), 도로와의 이격거리를 사용하였다. 산사태 발생에 영향을 미치는 인자를 자동화된 머신러닝인 AutoML을 실시하여 예측 성능이 뛰어난 XRT(extremely randomized trees)를 선정하였다. 변수 중요도 분석결과 지형적 요인 10개, 지질인자 9개, 사회적 영향성인 도로와의 이격 거리와 관련된 항목순으로 급경사지 불안정에 가장 많은 영향을 주는 것으로 분석되었다. 개발된 모델의 신뢰성 검증을 수행한 결과 AUC 83.977%의 예측율을 확보한 것으로 나타났다. 이 모델은 산사태 이력을 기반으로 한 현장조사 자료만을 이용하여 변수 중요도의 순위를 도출함으로써 그에 따른 산사태 발생 가능성을 확률적 및 정량적으로 평가하였다. 향후 의사 결정자들에게 현장조사를 통한 사면진단 안전평가 시 신뢰성 있는 근거를 제공하리라 판단된다.

키워드
AutoML
XRT
산사태 예측 모델
확률론적 예측
변수 중요도
References
1
Byun, L.S., Moon, H.K., 2010, The prediction of rock slope stability in Korea using artificial neural network, The Korean Society of Mineral and Energy Resources Engineers, 47(1), 31-44 (in Korean with English abstract).
2
Chae, B.G., Kim, W.Y., Cho, Y.C., Kim, K.S., Lee, C.O., Choi, Y.S., 2004, Development of a logistic regression model for probabilistic prediction of debris flow, The Journal of Engineering Geology, 14(2), 211-222 (in Korean with English abstract).
3
Dai, Z., Huang, Y., Cheng, H., Xu, Q., 2014, 3D numerical modeling using smoothed particle hydrodynamics of flow-like landslide propagation triggered by the 2008 Wenchuan earthquake, Engineering Geology, 180, 21-33.
10.1016/j.enggeo.2014.03.018
4
Di, B., Zhang, H., Liu, Y., Chen, N., Stamatopoulos, C., Luo, Y., Zhan, Y., 2019, Assessing susceptibility of debris flow in southwest China using gradient boosting machine, Scientific Reports, Published Online.
10.1038/s41598-019-48986-531467342PMC6715629
5
Galelli, S., Castelletti, A., 2013, Assessing the predictive capability of randomized tree-based ensembles in streamflow modelling, Hydrology and Earth System Sciences, 17(7), 2669-2684.
10.5194/hess-17-2669-2013
6
Geurts, P., Ernst, D., Wehenkel, L., 2006, Extremely randomized trees, Machine Learning, 63(1), 3-42.
10.1007/s10994-006-6226-1
7
Hong, W.P., Kim, W.Y., Song, Y.S., Lim, S.G., 2004, Prediction of landslide using artificial neural network model, Journal of the Korean Geotechnical Society, 29(8), 67-75 (in Korean with English abstract).
8
Kim, J.C., Lee, S.M., Jung, H.S., Lee, S.R., 2017, Landslide susceptibility mapping using random forest and boosted tree models in Pyeong-Chang, Korea, Geocarto International, Published Online, 1000-1015.
10.1080/10106049.2017.1323964
9
Merghadi, A., Yunus, A., Dou, J., Whiteley, J., ThaiPham, B., Bui, D.T., Avtar, R., Abderrahmane, B., 2020, Machine learning methods for landslide susceptibility studies: a comparative overview of algorithm performancem, Earth-Science Reviews, 207.
10.1016/j.earscirev.2020.103225
10
Nagarajah, T., Poravi, G., 2019, A review on automated machine learning (AutoML) systems, Proceedings of the 2019 5th International Conference for Convergence in Technology (I2CT), 1-6.
10.1109/I2CT45611.2019.9033810
11
Nam, K.H., Wang, F.W., 2019, The performance of using an autoencoder for prediction and susceptibility assessment of landslides: a case study on landslides triggered by the 2018 Hokkaido Eastern Iburi earthquake in Japan, Geoenvironmental Disasters, 6(19), 1-14.
10.1186/s40677-019-0137-5
12
Park, S.Y., Kim, J.S., 2019, Landslide susceptibility mapping based on random forest and boosted regression tree models, and a comparison of their performance, Applied Sciences, Published Online, 1-19.
10.3390/app9050942
13
Sassa, K., Nagai, O., Solidum, R., Yamazaki, Y., Ohta, H., 2010, An integrated model simulating the initiation and motion of earthquake and rain induced rapid landslides and its application to the 2006 Leyte landslide, Landslides, 7, 219-236.
10.1007/s10346-010-0230-z
14
Xia, B., Zhang, H., Li, Q., Li, T., 2015, PETs: A stable and accurate predictor of protein-protein interacting sites based on extremely-randomized trees, IEEE Transactions on Nanobioscience, 14(8), 882-893.
10.1109/TNB.2015.249130326529772
Information
  • Publisher :Korean Society of Engineering Geology
  • Publisher(Ko) :대한지질공학회
  • Journal Title :The Journal of Engineering Geology
  • Journal Title(Ko) :지질공학
  • Volume : 30
  • No :3
  • Pages :315-325
  • Received Date : 2020-08-26
  • Revised Date : 2020-09-17
  • Accepted Date : 2020-09-21
  • DOI :https://doi.org/10.9720/kseg.2020.3.315
Journal Informaiton The Journal of Engineering Geology The Journal of Engineering Geology
Online Submission
  • NRF
  • KOFST
  • KISTI Current Status
  • KISTI Cited-by
  • crosscheck
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close