Prediction of Forex Prices on USD/NGN Using Deep Learning (LSTM and GRU) Techniques

Abstract

The goal of the project is to develop a model to forecast the Foreign Exchange (FOREX) prices of United State Dollar to Nigerian Naira (USD/NGN), utilizing two machine learning algorithms, including Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU). These were chosen for this study because they have been found to be effective in previous studies that have been examined. The principles of machine learning and its applications, as well as the many machine learning techniques and algorithms will be covered in this study. Additionally, various extraction methods that will be used in the study will be presented. Data from the Investing.com dataset would be retrieved for this study's purpose and divided into training and test sets. Using the two machine learning techniques previously mentioned, the model would be trained and tested. Then, to measure the model's performance in terms of accuracy and precision, Mean Squared Error, Root Mean Squared Error, and Mean Absolute Error would be utilized. The results obtained showed that, GRU performed better than LSTM with a 0.950 Test R2 score and an adjusted R2 score of 0.122. The RMSE is way lower than LSTMs at 0.105 and MAE is even lower at 0.950.

Keywords

• forex, deep learning, long short term memory, gated recurrent unit.