The Long Short-Term Memory of GBP/CNY Exchange Rate Forecasts

Abstract

The issue of exchange rate forecasting has always been a hot topic, and with the increasingly close relationship between the UK and Chinese import and export trade, the GBP/CNY exchange rate has received increasing attention. Forecasting the GBP/CNY can help trading companies on both sides to effectively control their risks and help foreign exchange investors find arbitrage opportunities from it. For forecasting methods of foreign exchange, the main methods include traditional time series methods and deep learning methods. This study mainly uses the LSTM model to forecast the exchange rate of GBP/CNY from 31 January 2020 to 30 September 2021 and compares the results with ARIMA and GRU models, using RMSE, MAE and MAPE as the evaluation index of the results. Based on the results of the LSTM, ARIMA, GRU models, the RMSE values were 0.04268, 0.043791, 0.051312 respectively, we found that that the LSTM model has the best short-term forecasting results for GBP/CNY. We argue that the main reason for the LSTM to be the best model is that ARIMA and GRU models are more susceptible to parameter effects due to the shortcomings of the models themselves.