RBF神经网络在进出口总额预测中的应用

Application of RBF Neural Network in forecast of Import and Export Total

径向基函数(RBF)神经网络隐含层R由一组径向基函数构成。其执行通过非线性基函数的线性组合,从Rn到Rm的非线性变换。模型设计含数值归一化、样本数据预处理、模拟预测结果及分析。在进出口总额预测的应用证明,RBF网络可在数值上有效逼近时间序列内难以定时描述的相互关系,优于非线形回归和BP网络预测。

The hidden-layer R of RBF neural network consists of a set of RBF (radial base function). It is carried out through the linear association of non-linear base function, and the non-linear transformation from Rn to Rm. Model design includes number normalization, sample data preprocessing, simulating and analyzing forecasted result. The application of the import and export total prediction shows that RBF network approaches the interrelation effectively which is hard to time described in time series, and it is superior to non-linear regression and BP network in prediction.