基于局部线性嵌入与主成分分析的优化的RBF降水预报模型  

The Optimized RBF for Rainfall Forecasting Based on Locally Linear Embedding and Principal Component Analysis

作　　者：蒋林利[1] 

机构地区：[1]柳州师范高等专科学校数学与计算机科学系,广西柳州545004

出　　处：《河池学院学报》2013年第2期39-44,共6页Journal of Hechi University

基　　金：广西教育厅科研项目(201204LX506);柳州师专科研项目(LSZ2011B007)

摘　　要：针对用主成分分析算法对非线性数据降维,存在丢失局部结构数据信息的不足,提出了将主成分分析与局部线性嵌入相结合的正交投影降维算法。首先采用主成分分析方法降维可算出特征空间矩阵,然后采用局部线性嵌入方法降维后提取特征空间矩阵,再将这两种方法获得的特征空间矩阵正交投影进行降维的主要信息作为预报因子,最后利用优化的径向基函数RBF建立网络模型,以此对广西5月区域日降水量进行降水预报,结果显示,该模型在预报性能上优于同期的T213日降水预报。This paper, aiming at using principal component analysis algorithm for nonlinear data dimen- sionality reduction and the deficiencies of a lost local structure information, puts forward a kind of method called the orthogonal projection algorithm for dimensionality reduction which combines the two methods of PCA and LLE. The paper firstly introduces the method of PCA of dimensionality reduction to calculate special space matrix, and the method of LLE to calculate special space matrix ; and then makes the main information of the special space matrix of the orthogonal projection algorithm for dimensionality reduction to be the forecast factors; finally, establishes a net- work model by using the optimized radial basis function so as to forecast daily precipitation rainfall of Guangxi in May. The result of rainfall forecasting shows that this model is superior to T213 of the same period in the aspect of the prediction function.

关 键 词：局部线性嵌入 主成分分析 RBF神经网络 降维 预测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] O242[自动化与计算机技术—控制科学与工程]