基于支持向量机模型的滑坡危险性评价

Landslide Susceptibility Assessment Based on Support Vector Machine Model

以长安区为研究区,在分析研究区地质环境条件以及滑坡发育特征的基础上,选取12类因子作为评价指标,通过Spearman对各因子之间的相关性进行分析;分别采用线性核函数(LN)、多项式核函数(PL)、径向基核函数(RBF)、Sigmoid核函数下的支持向量机模型对研究区地质灾害危险性进行评价,利用ArcGIS软件生成最终的危险性评价结果图,将评价结果图划分为四个危险等级,分别为极高危险区、高危险区、中危险区、低危险区;不同核函数下的支持向量机模型经ROC曲线验证得到:RBF-SVM模型的预测精度最高,此核函数下的支持向量机模型更适应于该研究区滑坡灾害危险性评价中。研究结论可以为类似地质环境条件下区域地质灾害危险性评价模型的选取提供参考。

Chang'an district is taken as the research area.Based on the analysis of geological environment conditions and the characteristics of landslide development in the study area,12 types of evaluation index factors are selected.Spearman was adopted to analyze the correlation among various factors.The SVM model with four types of kernel classifiers such as linear(LN),polynomial(PL),radial basis function(RBF),sigmoid(SIG)were used to the susceptibility assessment of the landslide hazard.The final susceptibility assessment result map was generated by ArcGIS software.The assessment result map was grouped into four classes,namely,low,moderate,high,and very high.The SVM assessment model under different kernel functions was verified by ROC curve:the prediction accuracy of RBF-SVM model is the highest,which is more suitable for the evaluation of landslide hazard in the study area.This conclusion can provide reference for the selection regional geological hazard assessment model under similar geological environment conditions.