改进的学习向量量化滑坡易发性评价模型研究  

Study on improved learning vector quantization landslide vulnerability evaluation model

作　　者：赵晓东 刘福 杨华 张泰丽[2] ZHAO Xiaodong;LIU Fu;YANG Hua;ZHANG Taili(College of Civil Engineering&Architecture,Dalian University,Dalian,Liaoning 116622,China;Nanjing Center,China Geological Survey,Nanjing 210016,China)

机构地区：[1]大连大学建筑工程学院,辽宁大连116622 [2]中国地质调查局南京地质调查中心,南京210016

出　　处：《测绘科学》2023年第5期239-246,共8页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(51374046)。

摘　　要：针对传统的学习向量量化模型只能进行欧式空间的度量问题,该文将在学习向量量化(LVQ)模型的基础上引入径向基核函数(RBF)建立径向基函数的学习向量量化(RBF-LVQ)评价模型。以文成县为研究区,结合GIS技术选取坡度、坡向、坡形、断层距离、地质岩组、极端小时降雨量、地形湿度指数、地表覆盖、风化层厚度、黏聚力10个评价因子构建滑坡易发性评价体系,随机选取70%数据作为训练样本,分别采用RBF神经网络、LVQ神经网络和RBF-LVQ模型进行滑坡灾害易发性评价,并将剩余的30%数据利用ROC曲线进行精度检验。结果显示,训练后的RBF-LVQ模型AUC值为0.88,优于RBF神经网络的0.85和LVQ的0.86。RBF-LVQ模型拥有更好的预测能力,可为研究区域提供模型和决策支持。Aiming at the problem of the traditional learning vector quantization(LVQ)model can only measure Euclidean space,the learning vector quantization based on radial basis function(RBF-LVQ)evaluation model was established by introducing the radial basis kernel function(RBF)on the basis of the learning vector quantization(LVQ)model in this paper.Taking Wencheng County as the research area,ten evaluation factors such as slope,slope direction,slope shape,fault distance,geological rock group,extreme hourly rainfall,terrain humidity index,surface coverage,weathered layer thickness and cohesion are selected to construct the landslide vulnerability evaluation system by combining GIS technology.70%of the data are randomly selected as training samples,and RBF neural network,LVQ and RBF-LVQ models are respectively used to evaluate the landslide vulnerability,the remaining 30%data is used for precision test by ROC curve.The results show that the AUC value of the trained RBF-LVQ model is 0.88,which is better than 0.85 of the RBF model and 0.86 of the LVQ model.The RBF-LVQ coupling model has better predictive ability,which can provide model and decision support for landslide vulnerability assessment in the study area.

关 键 词：滑坡 易发性评价 径向基函数 学习向量量化 

分 类 号：X915.5[环境科学与工程—安全科学]