基于机器学习的遵义市滑坡灾害易发性评价研究

Study on Landslide Susceptibility Assessment Based on Machine Iearning in Zunyi City

滑坡灾害是我国经常发生的地质灾害之一,滑坡灾害易发性评价可以从空间和概率上将研究区划分成不同的滑坡风险等级。滑坡易发性评价结果可为滑坡治理和防治做出重要的决策支持。本研究以贵州省遵义市作为研究区,选取高程、坡度、NDVI等八个影响因素作为本次滑坡易发性评价的滑坡影响因子,选用决策树、随机森林、GDBT三种机器学习算法作为滑坡易发性评价的训练模型。通过自然间断法将评价结果按照易发性大小分成极高易发区、高易发区、中易发区、低易发区和极低易发区五类。使用ROC曲线和滑坡点密度对三种模型的效果进行对比分析。

Landslide disaster is one of the most common geological disasters in China.The evaluation of landslide disaster susceptibility can be divided into different risk levels from spatial and probability.The evaluation results of landslide susceptibility can provide important decision support for landslide control and prevention.In this paper,Zunyi City of Guizhou Province is taken as the research area,eight influencing factors such as elevation,slope and NDVI are selected as the influencing factors of landslide susceptibility evaluation,and three machine learning algorithms including decision tree,random forest and GDBT are selected as the training models of landslide susceptibility evaluation.Natural discontinuity method was used to classify the evaluation results into five types:extremely high,high,medium,low and extremely low.ROC curve and landslide point density were used to compare and analyze the effects of the three models.