A Deep Learning Model with Long Short-Term Memory (DLSTM) for Prediction of Currency Exchange Rates

Abstract

The objective of this research is to implement a Deep Learning model with Long Short-Term Memory (DLSTM) for prediction of the currency exchange rate. The system predicts the currency exchange rate using a fusion of data from the following financial inputs: gross domestic product rate (GDP), interest rate, inflation rate, balance account and trade balance as well as a finite set of previous exchange rates. We have evaluated the model performance by considering the currency exchange rates of the Thai Baht to the US dollar using historical data from the Bank of Thailand for ten years, from April 2009 to April 2019. To evaluate the effectiveness of the DLSTM model, we have considered the mean square error (MSE) and the mean absolute percentage error (MAPE). The best results have shown that the DLSTM model leads to a very low error value for the MSE and the MAPE at 0.0027 and 0.2844, respectively. We have compared the proposed DLSTM model prediction performances with those of the NARX neural network model; our research results show the obvious advantage of the proposed DLSTM model.