Abstract:
—Exchange rate forecasting is a vital problem in the
economic aspect of every country in the world. Prediction of the
foreign exchange rate is a very complex and challenging task. A
more in-depth analysis and forecasting techniques assist the traders
in good decision-making in their commercial activities. This paper
discusses forecasting of USD to LKR foreign exchange rate using
Artificial Neural Network (ANN) and Recurrent Neural Networks
(RNN). This study used two variant Recurrent Neural Networks,
Long Short Term Memory (LSTM) and Gated Recurrent Unit
(GRU). Rectified Linear Unit (ReLU) is used as an activation
function. Adam and Stochastic Gradient Descent (SGD) are used
as the optimizers in this research. The study mainly compares the
performance of ANN, LSTM, and GRU prediction rates with two
different optimizers Adam and SDG. Mean Square Error (MSE) is
used as the loss function. The study finds that GRU with Adam
optimizer performs better than other approaches in terms of R2
squared (Coefficient of determination), Root Mean Squared Error
(RMSE), Mean Absolute Error (MAE). In contrast, LSTM performs
better with SDG optimizer when compared to Adam.
