基于不同机器学习模型的滑坡易发性分析及适应性评估
DOI:
作者:
作者单位:

江苏省南京市南京信息工程大学水文与水资源工程学院

作者简介:

通讯作者:

中图分类号:

P642.22;X43

基金项目:

河北省省级科技计划资助项目(19275408D);江苏省水利科技项目(2020040);国家自然科学基金面上项目(41671022;41877158);江苏省研究生科研与实践创新计划项目(KYCX23_1375)


Landslide Susceptibility Analysis and Adaptability Evaluation Based on Different Machine Learning Models
Author:
Affiliation:

College of Hydrology and Water Resources Engineering, Nanjing University of Information Technology, Nanjing, Jiangsu Province

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    机器学习模型因其强大的特征提取能力被广泛应用于滑坡易发性评价,在应用中其算法在不断改进。为解决常见机器学习模型中精度不高的问题,本文将分组卷积神经网络模型(GCNN)引入滑坡易发性评价,并与多种常见机器学习模型结果进行对比分析,综合评估不同机器学习模型在滑坡易发性评价的适应性。以河北省为研究区,从致灾因子、孕灾环境、承灾体这个三个方面出发,共选取18个影响因子,选择GCNN模型和目前常见的机器学习模型——卷积神经网络模型(CNN)、逻辑回归模型(Logistic)、随机森林算法模型(RF)和支持向量机模型(SVM)构建出相应的易发性评价模型,将研究区划分为五类滑坡易发性区域,并对区划的精确性进行综合评价。研究表明,与其他四种机器学习模型相比,GCNN模型在各混淆矩阵指标下拥有更高评分,更适合进行滑坡易发性区划,得到的滑坡易发区划结果与实际发生滑坡点的一致性较好,划分的滑坡灾害易发区更加准确。

    Abstract:

    Machine learning models are widely used for landslide susceptibility evaluation due to their powerful feature extraction ability, and their algorithms are constantly improving in application. To address the issue of low accuracy in traditional machine learning models, this paper introduces the group convolutional neural network (GCNN) model into landslide susceptibility evaluation, and compares and analyzes the results with various traditional machine learning models to comprehensively evaluate the applicability of the GCNN model in landslide susceptibility evaluation.Taking Hebei Province as the study area, from three aspects of disaster causing factors, disaster pregnant environment and disaster bearing body, a total of 18 influencing factors were selected, GCNN and the current mainstream machine learning models - convolutional neural network model (CNN), logical regression model (Logistic), random forest algorithm model (RF) and support vector machine model (SVM) were selected to build the corresponding vulnerability assessment model, divide the research area into five types of landslide prone areas, and comprehensively evaluate the accuracy of the zoning.Research has shown that compared with the other four machine learning models, GCNN has higher scores under various confusion matrix indicators and is more suitable for landslide susceptibility zoning; The results of landslide prone zoning obtained by GCNN are consistent with the actual landslide occurrence points, and the landslide prone areas classified are more accurate.

    参考文献
    相似文献
    引证文献
引用本文

王洁,林诚杰,梁峰铭,等. 基于不同机器学习模型的滑坡易发性分析及适应性评估[J]. 科学技术与工程, , ():

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2024-04-17
  • 最后修改日期:2024-05-08
  • 录用日期:2024-05-22
  • 在线发布日期:
  • 出版日期:
×
亟待确认版面费归属稿件,敬请作者关注