基于局域支持向量机的旱涝预测  

Drought and Flood Prediction Research Based on Local Support Vector Machine

作　　者：李涛[1] 李娟[1] 

机构地区：[1]南京信息工程大学电子信息工程学院,南京210044

出　　处：《湖北农业科学》2015年第8期1991-1994,共4页Hubei Agricultural Sciences

基　　金：国家公益性行业(气象)科研专项(GYHY201306070)

摘　　要：降水量的变化受到许多因素影响,其动态特征呈现复杂的非线性,使得预测难度较大。为了提高降水量预测精度,提出了一种基于局域支持向量机的降水量预测方法,对月降水量时间序列进行参数提取,构造相空间,使用支持向量回归模型代替局域线性模型,使用邻近点训练该局域支持向量回归模型。仿真结果表明,该方法预测精度高,在旱涝预测方面有较好的应用前景。The research forecast monthly rainfall has significant meaning for the national production, disaster prevention and mitigation. The change of precipitation is influenced by many factors, the rainfall present complicated nonlinear dynamic characteristics, making prediction is difficult. In order to improve the prediction accuracy, a precipitation prediction method based on local support vector machine was proposed, parameters of monthly precipitation time series were extracted first, constructing the phase space, using support vector regression model instead of local linear model, with neighboring points in training the local support vector regression model. The simulation results showed that the method was of high precision, had good application prospect in the drought and flood prediction.

关 键 词：降水量预测 相空间 局域预测 支持向量机 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]