Study on Exchange Rate Volatility under Cross-border RMB Settlement Based on Multi-layer Neural Network Algorithm

Abstract

In order to increase profits, foreign trade enterprises need to reduce costs. But cross-border RMB settlement can reduce costs of foreign trade enterprises, which avoids exchange rate risks to some extent and reduces losses. However, cross-border RMB settlement will still be affected by exchange rate changes. In order to explore the law of exchange rate changes and make predictions to reduce the impact of exchange rate changes, the multi-layer neural network algorithm was used to train and test the exchange rates of the USD, EUR, JPY and HKD between November 2017 and July 2018 on the Matlab. The result indicated that the change of currency exchange rate was regular, and different currencies have different characteristics of change. The multi-layer neural network algorithm could accurately predict the exchange rate changes of most currencies and had the best performance in predicting the exchange rates of the USD and EUR, especially the EUR and the second best performance in predicting the exchange rate of the HKD; it could predict the general trend though it had the poorest performance in predicting the exchange rate of the JPY.