摘要
为了改进神经网络的预测性能,更精确地预测人民币汇率,提出一种新的汇率时间序列预测方法,即利用基于经验模态分解(EMD)的Elman网络进行预测.首先对人民币兑美元的汇率序列做了非线性检验和非平稳性检验,然后对该序列进行经验模态分解,将得到的固有模态函数作为神经网络的输入变量,并在确定神经网络的关键参数后进行预测.实证结果表明,利用基于EMD的Elman网络进行人民币汇率预测能够取得更好的效果.
In order to improve the forecasting performance of neural networks and to forecast the RMB exchange rate more accurately, a new method for exchange rate time series forecasting was proposed. That is empirical mode decomposition （EMD） based Elman neural network ensemble learning paradigm. First, a non-linear and non-stationary test was done to time series of RMB exchange rate against the U.S. dollar. Then, we decomposed the series into several Intrinsic Mode Functions by EMD, which were the input variables of the neutral network, determined the key parameters and did the forecasting. The empirical results have shown that the method proposed is more accurate and effective.
出处
《湖南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2009年第6期89-92,共4页
Journal of Hunan University:Natural Sciences
基金
国家社会科学基金重点资助项目(07AJL005)
高等学校博士学科点专项科研基金资助项目(20070532091)
全国高等学校青年教师奖励基金资助项目(教人司2002[123])
关键词
时间序列
汇率预测
经验模态分解
ELMAN网络
time series analysis
exchange rate forecasting
empirical mode decomposition （EMD）
Elman neutral network
