The Application of Improved Elman Neural Network in the Exchange Rate Time Series

In this paper, we select the Elman neural network method to improve because of its good non-linear effect of disturbance elimination, and present a new exchange rate time series prediction method. We point out a new improved Elman neural network model firstly, and then predict the time series of RMB exchange rate against U. S. dollar. Through the forecasting process, we determine the input variables for the network structure, and determine the neural network´s critical parameters to forecasting. The results show that the improved Elman network can obtain better results during the forecasting process.