Prediction of USD/RMB Price Change Based on BP Neural Network

Abstract

The market position of the RMB has gradually increased, and the price change of the US dollar against the RMB has become increasingly important. Therefore, it is vital to use scientific means to forecast the exchange rate forecasting model. In this thesis, the BP neural network model is established by MATLAB, and the price change of the USD/RMB exchange rate is analyzed and processed according to the characteristics of the exchange rate prediction model. Part of the data is elected for the network training, the other part is used to verify the accuracy of neural network prediction. The results indicate that the fitting accuracy is 94.6%. Continuing to modify some arguments of the model including changing the number of training rounds to 100000, and changing the training target to 1e-15 with the learning rate remaining unchanged, The equal accuracy could be further increased to 96%. In the results, the model can precisely, mirror the trend of fluctuations, verify the effectiveness of this forecasting method, provide an important reference for financial institutions and personnel engaged in the financial industry, and have great significance for foreign exchange management to effectively avoid market risks and cope with the impact of exchange rate fluctuations.