EXPERT: EXchange Rate Prediction Using Encoder Representation from the Transformers

In this study, we introduce a Transformer-based forecasting tool termed EXPERT (EXchange rate Prediction using Encoder Representation from the Transformers) and apply it to the domain of exchange rate forecasting. To achieve this, we first developed and trained the Transformer-based forecasting model; we then evaluated its performance on a diverse set of 9 currency pairs with various characteristics. Finally, we benchmarked its effectiveness against six established and respected forecasting models: Linear Regression, Random Forests, Stochastic Gradient Descent, XGBoost, Bagging Regression, and Long Short-Term Memory. Our dataset covers the period from 1999 to 2022. The models were evaluated for their ability to predict the next day's closing price using three performance metrics. In addition, the EXPERT system was evaluated for extending forecast horizons and as the core of a trading strategy. The universal robustness of the proposed model was tested using the Multiple Comparisons with the Best (MCB) on five samples of our dataset.