ISSN 1003-8035 CN 11-2852/P
    陈国圳,刘军旗,崔婧慧,等. 不同机器学习模型在湖北巴东滑坡易发性评价中的应用[J]. 中国地质灾害与防治学报,2024,35(0): 1-12. DOI: 10.16031/j.cnki.issn.1003-8035.202310011
    引用本文: 陈国圳,刘军旗,崔婧慧,等. 不同机器学习模型在湖北巴东滑坡易发性评价中的应用[J]. 中国地质灾害与防治学报,2024,35(0): 1-12. DOI: 10.16031/j.cnki.issn.1003-8035.202310011
    CHEN Guozhen,LIU Junqi,CUI Jinghui,et al. Application of different machine learning models in landslide susceptibility assessment in Badong County, Hubei province[J]. The Chinese Journal of Geological Hazard and Control,2024,35(0): 1-12. DOI: 10.16031/j.cnki.issn.1003-8035.202310011
    Citation: CHEN Guozhen,LIU Junqi,CUI Jinghui,et al. Application of different machine learning models in landslide susceptibility assessment in Badong County, Hubei province[J]. The Chinese Journal of Geological Hazard and Control,2024,35(0): 1-12. DOI: 10.16031/j.cnki.issn.1003-8035.202310011

    不同机器学习模型在湖北巴东滑坡易发性评价中的应用

    Application of different machine learning models in landslide susceptibility assessment in Badong County, Hubei province

    • 摘要: 中国是世界上发生滑坡灾害最频繁的国家之一,滑坡易发性评价有助于防灾减灾工作。由于不同机器学习模型在不同区域的适配程度不同,为更好开展湖北省巴东县的滑坡灾害防治工作,选取坡度、坡向、曲率、起伏度、地层、覆盖层、NDVI、道路密度、水系密度、斜坡结构10个影响因子,采用逻辑回归(LR)、支持向量机(SVM)、多层感知机(MLP)和随机森林(RF)四种模型进行滑坡易发性评价。并通过受试者工作特征曲线(ROC)、均方误差与决定系数等指标、滑坡-研究区占比三种评价方式用于评价模型精度。实验结果表明,不同模型在不同评价方式中存在差异,但总体而言,RF模型精度最高且绘制出的易发性分区图更合理。四个模型绘制的易发性区域分布图相似,极高易发区和高易发区主要分布于南边沿江地区,西南沿岸的官渡口镇、焦家湾村等附近地区表现出较高易发性,该评价结果可以为巴东县的滑坡治理提供参考。

       

      Abstract: China is one of the countries most frequently affected by landslide disasters in the world, making landslide susceptibility assessment crucial for effective disaster prevention and mitigation. Due to variations in the adaptability of different machine learning models in different regions, in order to better carry out landslide disaster prevention and control work in Badong County, Hubei Province, ten influencing factors including slope gradient, slope direction, curvature, degree of undulation, stratigraphy, overburden, NDVI, road density, water system density, and slope structure were selected. Four different models, including Logistic Regression (LR), Support Vector Machine (SVM), Multilayer Perceptron (MLP), and Random Forest (RF), were used for landslide susceptibility evaluation. Three evaluation methods were used to assess the accuracy of the model: Receiver Operating Characteristic (ROC) curves, mean square error, determination coefficient, and the ratio of landslide to study area. The experimental results show that there are differences among the models in different evaluation methods. Overall, the RF model exhibits the highest accuracy and generates more reasonable susceptibility zoning maps. The susceptibility distribution maps generated by the four models are similar, with high and very high susceptibility areas predominantly located in the southern riverside area. Areas near Guandukou Town and Jiaojiawan Village along the southwest coast exhibit relatively high susceptibility. The assessment results can provide reference for landslide control in Badong County.

       

    /

    返回文章
    返回