Comparison of Forecasting Foreign Exchange Rate and Profitability of Trading Strategy using ARIMA and GARCH Model : Cases of AUD/USD, EUR/USD, GBP/USD, USD/CAD and USD/CHF

Abstract

This study conducts the comparison of efficiency in forecasting of 5 currency pairs, AUD/USD (Australian Dollar / US. Dollar), EUR/USD (Euro / US. Dollar), GBP/USD (Great Britain Pound / US. Dollar), USD/CAD (US. Dollar / Canadian Dollar), USD/CHF (US. Dollar / Swiss Franc) rates forecast by using ARIMA and GARCH model. We consider both size forecasting and direction forecasting to build the trading strategy and compare return. Then apply direction forecasting to trading strategy using daily close price data during September 2015 – March 2019. Data are separate into 2 parts, the first part used as In sample data, September 2015 – 2018 for 3 years, to find the optimal model of forecasting and the optimal stop loss of each model. The second part used as Out of sample data, October 2018 – march 2019, to determine the efficiency of model. The work described in this paper, the accuracy of size forecasting is compared with Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) show that GARCH model has more accuracy than ARIMA model for almost every currency pairs except AUD/USD, while ARIMA outperform in direction forecasting except USD/CHF when compared with win rate. After applying direction forecasting into trading strategy which is long and short position in 1 day with the optimal stop loss. The results show that ARIMA model is capable to earn more profit and has less drawdown than GARCH model in all currency pairs.