Search for a command to run...
The foreign exchange (FOREX) market is one of the biggest financial markets in the world. More than 5.1 trillion dollars are traded each day in the FOREX market by banks, retail traders, corporations, and individuals. Due to complex, volatile, and high fluctuation, it is quite difficult to guess the price ahead of the actual time. Traders and investors continuously look for new methods to outperform the market and to earn a higher profit. Therefore, researchers around the world are continuously coming up with new forecasting models to successfully predict the nature of this unsettled market. This paper presents a new model that combines two powerful neural networks used for time series prediction: Gated Recurrent Unit (GRU) and Long Short Term Memory (LSTM), for predicting the future closing prices of FOREX currencies. The first layer of our proposed model is the GRU layer with 20 hidden neurons and the second layer is the LSTM layer with 256 hidden neurons. We have applied our model on four major currency pairs: EUR/USD, GBP/USD, USD/CAD, and USD/CHF. The prediction is done for 10 minutes timeframe using the data from January 1, 2017 to December 31, 2018, and 30 minutes timeframe using the data from January 1, 2019 to June 30, 2020 as a proof-of-concept. The performance of the model is validated using MSE, RMSE, MAE, and R2 score. Moreover, we have compared the performance of our model against a standalone LSTM model, a standalone GRU model and simple moving average (SMA) based statistical model where the proposed hybrid GRU-LSTM model outperforms all models for 10-mins timeframe and for 30-mins timeframe provides the best result for GBP/USD and USD/CAD currency pairs in terms of MSE, RMSE, and MAE performance metrics. But in terms of R2 score, our system outperforms all compared models and thus proves itself as the least risky model among all.