Statistical Forecasting of Exchange Rates: A Review A Comparison of Standard Time Series Models and Machine Learning Technique

Forecasting exchange rates is a vital component of the global financial system, influencing international trade, market investments, and monetary policy. This thesis investigates the performance of traditional time series models and machine learning approaches in the predicting of features of exchange rate fluctuations, such as, price level, returns, direction, etc. (think about other different features that you could be interested in comparing) with an emphasis on identifying the contexts in which each methodology is most effective. The study will test formally the accuracy of machine learning techniques and time series models at capturing patterns within exchange rate data, in terms of various evaluation metrics The results of this research contribute to the understanding of how different forecasting methods perform under varying conditions. Insights gained are intended to aid practitioners in finance, economics, and policymakers in the selection of the most suitable tools for enhancing prediction accuracy, and thereby supporting better strategic decisions.