Exchange Rate Prediction with Machine Learning, Deep Learning, and Time Series Methods Using Alternative Data

Abstract

Using alternative data to predict macroeconomic variables is efficient and consumes less time. This study aims to find the effectiveness of using alternative data such as the NASDAQ Index, NIFTY 50 Index, and SENSEX Index to forecast Exchange rates. The study used USD conversion rate data from various websites such as money control, yahoo finance, India stat, official Reserve Bank of India (RBI) website. The experiment-based research uses Machine Learning (ML), Deep Learning (DL), and Time Series Modeling to predict conversion rates. The study reveals that the NASDAQ index significantly affects conversion rate, whereas the NIFTY 50 and SENSEX indexes had less impact. It is evident from this study that the Ensemble ML model gives the best prediction results with 90% accuracy. DL models were unreliable, and time series forecasting gave considerable accuracy.