The Analysis of Global RMB Exchange Rate Forecasting and Risk Early Warning Using ARIMA and CNN Model

Abstract

The purposes are to predict exchange rate fluctuations more accurately and enhance Chinese enterprises’ ability to avoid exchange rate risks. Renminbi (RMB) exchange rate fluctuation’s prediction methods are studied based on data mining technology. The Auto-Regressive Integrated Moving Average (ARIMA) model is introduced first using a modeling method that combines linear and nonlinear models. The linear prediction is obtained by the ARIMA model’s application in the RMB exchange rate’s dynamic fluctuation analysis. The nonlinear residual prediction is obtained by integrating the ARIMA model with the convolutional neural network (CNN) algorithm. The RMB exchange rate fluctuations’ influence mechanism on China’s economic growth is explored by theoretical analysis and empirical research. The US dollar’s daily central parity rate (USD) data against the RMB from September 2015 to March 2019 are selected for model verification, obtaining the exchange rate’s logarithmic return sequence (RUSD).