摘要
选取自1994—2012年来广西共87个暴雨致洪灾害过程,以暴雨时间长度、暴雨过程降水极值、暴雨过程降水均值三个指标作为致灾源因子,以暴雨洪涝造成的直接经济损失作为主要灾情因子,通过对数变换获得更加光滑的灾情因子序列.利用统计学理论建立支持向量回归模型,采用网格搜索法进行SVM的参数寻优,利用选取的致灾源因子对灾情序列进行回归预测.结果表明,基于支持向量机的回归预测模型,其拟合优度、拟合效果均优于传统的逐步回归.该研究为探讨暴雨致灾源因子与灾情因子的非线性关系提供了一条新的途径.
This article focuses on a total of 87 heavy rain flooding in Guangxi from the year of 1994 to 2012,with heavy rain time length,average precipitation extreme precipitation,rainstorm process heavy rain process,the three indicators as a source of hazard -formative factors,with heavy rain floods that cause a direct economic loss as maj or disaster factors,by logarithmic transformation in order to obtain more slippery disaster factor sequence.Support vector regression model is estab-lished by using statistics theory,using the grid search method for SVM parameters optimization,u-sing the selected source factor to cause return proj ections for disaster sequence.The results show that the regression forecasting model based on support vector machine (SVM),the goodness of fit are su-perior to the traditional regression fitting effect.This study is to explore rainstorm nonlinear relation-ship of the source factors and disaster factors,thus provides a new way.
出处
《广西师范学院学报（自然科学版）》
2014年第3期31-35,42,共6页
Journal of Guangxi Teachers Education University(Natural Science Edition)
基金
广西科学研究与技术开发项目(1355010-8)
广西自然科学基金(GXNSFAA019280)
关键词
支持向量机
统计学理论
对数变换
逐步回归
广西
暴雨
SVM
statistical theory
logarithmic transformation
stepwise regression
Guangxi
heavy rain
