节点文献

基于粒子群优化支持向量机的延长县滑坡易发性评价

Evaluation of Landslide Susceptibility in Yanchang County Based on Particle Swarm Optimization-Based Support Vector Machine

  • 推荐 CAJ下载
  • PDF下载
  • 不支持迅雷等下载工具,请取消加速工具后下载。

【作者】 郭天颂张菊清韩煜钟炎伶谭锦蓉韦建成

【Author】 Guo Tiansong;Zhang Juqing;Han Yu;Zhong Yanling;Tan Jinrong;Wei Jiancheng;School of Geological Engineering and Geomatics, Chang′an University;State Key Laboratory of Geographic Information Engineering, Chang′an University;

【通讯作者】 张菊清;

【机构】 长安大学地质工程与测绘学院长安大学地理信息工程国家重点实验室

【摘要】 参数优化问题直接影响着支持向量机的预测精度和泛化能力,粒子群优化算法具有全局最优搜索能力,因此通过粒子群算法优化支持向量机参数可以有效提高预测精度。以延长县历史滑坡数据为基础,分析了岩性、地貌类型、土壤厚度、坡度、坡向、坡高与滑坡分布的关系,并利用滑坡密度值对各定性或定量因子进行了归一化处理;在此基础上,通过区域内所划分的16 300个斜坡单元作为评价单元,采用粒子群优化支持向量机(PSO-SVM)算法完成了延长县滑坡的易发性评价。从滑坡密度指标角度来看,评价结果中高易发区和极高易发区的历史滑坡数占比72.19%,通过滑坡面积百分比(LAR)等指标进行了有效的验证,均显示出对滑坡易发性评价效果良好。

【Abstract】 The parameter optimization directly affects the prediction accuracy and generalization of SVM. The particle swarm optimization algorithm has global optimal search ability. Therefore, optimizing the support vector machine parameters by particle swarm optimization can effectively improve the prediction accuracy. Based on the historical landslide data of Yanchang County, this paper explores and analyzes the hazard factors of landslides, and uses ArcGIS platform to extract and analyze the relationship between lithology, landform, loess thickness, slope, slope direction, slope height and landslide distribution, and to use landslide density to normalize each qualitative or quantitative factor. On the basis of this, the PSO-SVM algorithm was used to evaluate the susceptibility of Yanchang County landslide through the area 16 300 divided by the slope unit as the evaluation unit. From the perspective of landslide density index, the number of historical landslides in the high-prone and ultra-high-prone areas account for 72.19%, validated by landslide percentage(LAR) and other indicators.Both show good susceptibility evaluation results.

【基金】 国家自然科学基金项目(41731066;41274004)
  • 【文献出处】 地质科技情报 ,Geological Science and Technology Information , 编辑部邮箱 ,2019年03期
  • 【分类号】P642.22
  • 【被引频次】20
  • 【下载频次】439
节点文献中: 

本文链接的文献网络图示:

本文的引文网络