基于改进的RBF神经网络的人民币汇率预测研究

Research on prediction of RMB exchange rate based on improved RBF neural network

针对RBF神经网络分段算法中对近似线性时间序列数据预测误差较大这一不足,在原有RBF神经网络模型基础上提出了一种改进算法。该算法以分段取中心值为基础,优化原算法中径向基函数中心点值的确定,提高了对近似线性时间序列数据预测的准确度。通过对近两年美元兑人民币汇率数据的预测测试,表明改进算法在预测准确性比原算法有较大提高。

In view of considerable data prediction errors of approximate linear time series data in partition algorithm of RBF neural network,a new improved algorithm is presented on the basis of original RBF neural network.This improved algorithm takes central value by each section as a basis,optimizes the determination of values of central point of original algorithm of radial base function and improves accuracy of data prediction of approximate linear time series data.Prediction experiment of exchange rate of American Dollar for RMB in recent two years has proved that the prediction accuracy of this improved algorithm is comparatively higher than that of original algorithm.