集对分析径向基函数神经网络预测模型  

Prediction Model of Radial Basis Function Neural Network Based on Set Pair Analysis

作　　者：陈晶[1,2] 王文圣[1,2] 李跃清[1] 

机构地区：[1]中国气象局成都高原气象研究所,四川成都610072 [2]四川大学水利水电学院,四川成都610065

出　　处：《水文》2011年第2期11-14,共4页Journal of China Hydrology

基　　金：科技基础性工作专项项目<我国水文学方法创新体系建设研究>(2009IM020100);新世纪优秀人才支持计划;中国气象局成都高原气象研究所基金项目(LPM2010007)

摘　　要：将集对分析与径向基函数神经网络结合,提出了集对分析径向基函数神经网络预测模型。模型思路是将研究对象t-1时和t时的影响因子集构造为集对并计算联系度,由联系度的同一度、差异度、对立度及研究对象t-1时的值为输入,研究对象t时的值为输出,构建径向基函数神经网络。以年径流预测为例研究表明,模型结构清晰、步骤明确、预测精度较高,为集对分析应用于水文预测提供了新思路。The proposed SPA-RBFNN prediction model is a combination of set pair analysis （SPA） and radial basis function neural network （RBFNN）. The idea of SPA-RBFNN, firstly sets the impact factors of research object in both t-1 and t period of time as a pair, and calculates the connection degree of the pair, then uses its calculated homology degree, difference degree and antinomy degree, along with the situation of research object in t-1 period of time as model input, the situation of research object in t period of time as model output, finally finishes the model establishment. The case study of annual runoff prediction shows that SPARBFNN prediction model is characterized by explicit structure, easy realization and good prediction ability. The model construction idea provides a new thinking for the application of SPA in solving the hydrological prediction problems.

关 键 词：集对分析 径向基函数神经网络 径流预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP183[自动化与计算机技术—控制科学与工程]