数据驱动自适应更新的斜坡地质灾害易发性评价系统  

A data driven self-adaptive update landslide susceptibility assessment system

作　　者：王丽丽[1] 苏程[1] 冯存均 左石磊 黄智才[1] 章孝灿[1] 

机构地区：[1]浙江大学空间信息技术研究所,浙江杭州310027 [2]浙江省地理信息中心,浙江杭州310012

出　　处：《岩石力学与工程学报》2016年第A01期3076-3083,共8页Chinese Journal of Rock Mechanics and Engineering

基　　金：浙江省地理国情普查支持项目(H20141865)~~

摘　　要：为了实现斜坡地质灾害易发性评价能够随着影响因子动态变化和灾害点数目不断增加的自适应更新,使得评价精度能够随着影响因子认识的深入和样本点的丰富不断升高,提出一种数据驱动的斜坡地质灾害易发性评价系统自适应更新机制,以影响因子处理与筛选、支持向量机以及正确率、Kappa系数和AUC三个评价指标为理论支撑,设计并实现斜坡地质灾害易发性评价系统。该系统由数据预处理、易发性评价模型构建、评价模型评估、地质灾害易发性评价等4个模块构成,以数据变化触发系统自适应更新,可以支持影响因子的动态更新、地质灾害样本点的实时更新以及评价模型的迭代更新,使得评价模型能动态升级以适应最新的区域环境特征和地质灾害特征,保证评价模型的准确性。在浙江省文成县的实验结果表明,数据驱动自适应更新的斜坡地质灾害易发性评价系统能对地质灾害进行有效评价,并随着数据的丰富,将有效提高评价精度。To maintain the susceptibility assessment model up-to-date and to improve the accuracy of assessment result, a data driven self-adaptive update mechanism in landslide susceptibility assessment system is proposed to accommodate changes of conditioning factors and increasing number of landslide inventories. Based on the theory of conditioning factor processing and selecting method, support vector machine and three evaluation indices （accuracy, kappa coefficient and A UC）, a landslide susceptibility assessment system has been designed and implemented. The system includes 4 modules： data preprocessing, landslide assessment model construction, model performance evaluation and landslides prediction. The self-adaptive update mechanism will be triggered by data changes. To fit the landslide assessment model for the latest regional environment and characteristics of landslides, the system will rebuild the training set and reconstruct the landslide assessment model once the conditioning factors and landslide inventories change or the assessment result cannot satisfy requirements. The experimental results in Wencheng County, Zhejiang Province, China show that the system can be updated automatically with changes of data to accommodate the latest environment and can improve the landslide susceptibility assessment result effectively as the abundance of conditioning factors and landslide inventories.

关 键 词：边坡工程 斜坡地质灾害 易发性评价 数据驱动 自适应更新 支持向量机 

分 类 号：P642.22[天文地球—工程地质学]