摘要
泥石流是一种常见的地质灾害,对人类的生命和财产都会造成极大危害。泥石流危险性预测是防灾减灾的关键。以白龙江流域泥石流为例,首先利用主成分分析对原始数据进行分析,提取主成分,然后采用交叉验证的方式抽取训练样本与预测样本进行多分类支持向量机预测,建立预测模型,对泥石流危险性等级进行分类预测。结果表明:基于5折交叉验证的主成分分析与多分类支持向量机预测模型准确率可达到90%,可为泥石流危险性预测的研究提供计算模型依据。
Debris flow is a kind of common geological disaster,which can cause great harm to human life and property.The prediction of debris flow risk is the key of disaster prevention and mitigation.In this paper,the debris flow in Bailong River Basin was taken as an example.Firstly,principal component analysis is used to analyze the original data and extract the principal components.Then,cross validation is used to extract training samples and prediction samples for multi-classification support vector machine prediction,and a prediction model is established to classify and predict the risk level of debris flow.The results show that the prediction accuracy of principle component analysis and multi-classification support vector machine based on 5-fold cross validation can reach 90%.It can provide the basis of calculation model for the prediction of debris flow risk.
作者
刘超
乔圣扬
LIU Chao;QIAO Sheng-yang(Hebei GEO University,Shijiazhuang 050031,China)
出处
《河北地质大学学报》
2021年第4期83-89,共7页
Journal of Hebei Geo University
基金
河北省自然科学基金项目(D2019403182)
河北省高等学校科学技术研究项目(QN2019196)。
关键词
泥石流
主成分分析
交叉验证
多分类支持向量机
危险性预测
debris flow
principal component analysis
cross validation
multi-classification support vector machine
hazard prediction
