- Title
- Landslide susceptibility assessment based on clustering analysis and support vector machine
- Creator
- Huang, Faming; Yin, Kunlong; Jiang, Shuihua; Huang, Jinsong; Cao, Zhongshan
- Relation
- Yanshi Lixue yu Gongcheng Xuebao / Chinese Journal of Rock Mechanics and Engineering Vol. 37, Issue 1, p. 156-167
- Publisher Link
- http://dx.doi.org/10.13722/j.cnki.jrme.2017.0824
- Publisher
- Zhongguo Kexueyuan Wuhan Yantu Lixue Yanjiuso / Chinese Academy of Sciences, Institute of Rock and Soil Mechanics
- Resource Type
- journal article
- Date
- 2018
- Description
- The non-landslide grid cells are selected randomly and/or subjectively when the machine learning models, such as the support vector machine (SVM), are used to calculate the susceptibility indexes of regional landslides. However, it is difficult to determine whether the randomly selected non-landslide grid cells are reasonable "non-landslide" with very low susceptibility. To overcome this drawback, a model based on the combined clustering analysis and SVM is proposed. Firstly, the neural network with self-organizing mapping (SOM) is proposed to automatically classify the landslide susceptibility of all the grid cells into five classes: very low, low, moderate, high and very high susceptibility. Then, the reasonable non-landslide grid cells are selected from the area of very low susceptibility. Finally, the SVM is used to calculate the indexes of landslide susceptibility based on the recorded landslide grid cells, the selected non-landslide grid cells and the environmental factors. The proposed SOM-SVM model is used to calculate the susceptibility indexes of landslide in Wanzhou district of Three Gorges Reservoir area. The calculated results with the SOM-SVM model are compared with the results from the single SVM model which selects the non-landslide grid cells randomly. The results show that the SOM-SVM model has higher success and prediction rates than the single SVM. It is thus concluded that the non-landslide grid cells selected by the SOM neural network are more reasonable than the non-landslide grid cells selected randomly.
- Subject
- slope engineering; landslide susceptibility; non-landslide grid cells; self-organizing mapping neural network; support vector machine
- Identifier
- http://hdl.handle.net/1959.13/1412335
- Identifier
- uon:36462
- Identifier
- ISSN:1000-6915
- Language
- CA
- Reviewed
- Hits: 3005
- Visitors: 2997
- Downloads: 0
Thumbnail | File | Description | Size | Format |
---|