基于信息增益比-支持向量机的泥石流易发性评价

Assessment of Debris Flow Susceptibility Based on Information Gain Ratio-Support Vector Machine

基于对云南省东川区泥石流易发性评价研究的目的,以泥石流多发地东川区为例,采用多重共线性分析和信息增益比(IGR)模型筛选指标因子,在此基础上利用支持向量机模型(SVM)构建泥石流易发性评价模型。通过GIS平台,以栅格单元为评价单元,使用自然断点法制作研究区泥石流易发性分为极高易发区、高易发性区、中易发区、低易发区和极低易发区5个等级。结果表明,指标因子经过优化后,模型的曲线下面积(AUC)值为0.9033,说明指标因子选取合理,模型构建可靠,研究结果可为山区泥石流防灾减灾提供科学依据。

Based on the purpose of evaluating the mudslide susceptibility in Yunnan Province,Dongchuan District,a mudslide prone area,was used as an example to screen the index factors by using multicollinearity analysis and information gain ratio(IGR)model,on which a mudslide susceptibility evaluation model was constructed by using support vector machine model(SVM).The mudflow susceptibility classification map of the study area was produced by using the natural breakpoint method through a GIS platform with raster cells as the evaluation unit to quantitatively analyze the mudflow susceptibility.The results showed that the area under the curve(AUC)value was 0.9033 after optimization of the index factors,indicating that the index factors were selected reasonably and the model was constructed reliably,and the research results could provide scientific basis for mudslide disaster prevention and mitigation in mountainous areas.