Chinese currency fixing model using multivariate linear learning

Abstract

In June 2017, the People’s Bank of China (PBOC) announced a change in the Chinese Yuan (CNY) exchange rate regime, transitioning it to a managed-floating currency over which the central bank has regulatory powers. The PBOC releases the reference USD/CNY exchange rate daily at 9:30 AM local time. The aim of this research project is to develop a prediction model that uses multivariate linear learning techniques to forecast this daily exchange rate setting. The defining characteristic of the model developed within this project is that it combines the two major forecasting techniques, i.e., the global macroeconomic indicators and the historical USD/CNY exchange rate values into one prediction function. This function automatically learns from the historical fixing rates while taking into account the shift in the global financial markets. The behaviour and performance of the prediction model was successfully validated using existing research in CNY exchange rate forecasting. The results were benchmarked against the forecasting accuracy of other models existing in the public domain. The model was then incorporated into a web application developed through the course of this project, that served as a real-time information portal for users interested in the foreign exchange market.