Comparative Analysis of Currency Exchange Rate Forecasting Models Using Machine Learning on the Example of BRICS Currencies

The BRICS countries play a significant role in shaping the international economic landscape. Forecasting currency exchange rates of BRICS countries holds particular importance due to the potential impact of their currency fluctuations on global trade, investments, and financial markets. This study conducts an analysis of currency exchange rate prediction models utilizing machine learning methodologies, with a specific focus on BRICS currencies. This study assesses the effectiveness of different machine learning algorithms in the context of forecasting currency exchange rates within the BRICS countries. Specifically, it examines the performance of DeepAR, a novel deep/machine learning algorithm created by Amazon that integrates econometric autoregressive principles with recurrent neural network architectures, and Temporal Fusion Transformer, an advanced forecasting algorithm introduced by Google that combines LSTM structures with transformer mechanisms. These algorithms are compared against the benchmark forecast model ARIMAX from the field of econometrics. By identifying the most effective machine learning algorithms for currency exchange rate prediction, this research aims to enhance forecasting techniques in financial time series analysis.