基于不确定贝叶斯分类技术的滑坡危险性评价  

Landslide hazards assessment based on uncertain Bayesian classification method

作　　者：毛伊敏[1,2,3] 张茂省[1] 程秀娟[1] 彭喆[3] 

机构地区：[1]国土资源部黄土地质灾害重点实验室,陕西西安710054 [2]长安大学地质工程与测绘学院,陕西西安710036 [3]江西理工大学应用科技学院,江西赣州341000

出　　处：《中国矿业大学学报》2015年第4期769-774,共6页Journal of China University of Mining & Technology

基　　金：国家自然科学基金项目(41362015;51164012);国土资源调查项目(1212011140005);江西省自然科学基金项目(20122BAB201045);国家高技术研究发展计划(863)项目(2012AA061901)

摘　　要：由于引起滑坡的诱发因素(降雨)值无法准确获取,应用概率统计思想,刻画雨量值;依据延安黄土滑坡成灾机理研究,确定坡高、坡度、坡向、坡型、岩土体结构和植被覆盖为滑坡成灾基本因素;提出了不确定贝叶斯分类挖掘技术构建区域滑坡危险性评价模型方法,达到提高区域滑坡危险性预测精度的目的.以延安市宝塔区为例,验证了运用该方法进行区域滑坡危险评价的可行性.结果表明:采用不确定贝叶斯分类模型对研究区进行危险性划分,评价结果和实际滑坡发育情况吻合,合理地反映区内滑坡灾害发育的总体特征;不确定贝叶斯分类方法的零阶矩曲线下面积值比朴素贝叶斯方法有所提高,证明该方法科学合理.As the inducing factor of landslide,the amount of rainfall cannot be accurately obtained,in this paper probability statistics method was applied to describe the precipitation capacity.According to mechanism study of Yan＇an loess landslide,slope height,angle,type,structure of rock soil mass and vegetation cover were identified as the basic inducing factors of landslide.Landslide hazard assessment models based on uncertain Bayesian classification mining technology was proposed to raise the landslide accuracy prediction.Then the feasibility of the method in regional landslide hazard assessment is tested in Baota district of Yan＇an city in Shaanxi province.Research results show that the evaluation results based on uncertain Bayesian landslide spatial models are consistent with actual landslide development,which can properly reflect the general characteristics of regional landslide evolution.Meanwhile,the AUC value of uncertain Bayesian algorithm is higher than that of the naive Bayesian＇s,indicating that the new method is scientific and reasonable.

关 键 词：不确定贝叶斯模型 滑坡 危险性评价 黄土 

分 类 号：TU473[建筑科学—结构工程]