Title: Machine learning-based forecasting of significant daily returns in foreign exchange markets

Authors: Firuz Kamalov; Ikhlaas Gurrib

Addresses: Faculty of Engineering, Canadian University Dubai, Dubai, United Arab Emirates ' Faculty of Engineering, Canadian University Dubai, Dubai, United Arab Emirates

Abstract: Financial forecasting has always attracted an enormous amount of interest among researchers in quantitative analysis. The advent of modern machine learning models has introduced new tools to tackle this classical problem. In this paper, we apply machine learning algorithms to a hitherto unexplored question of forecasting instances of significant fluctuations in currency exchange rates. We carry out an extensive comparative study of ten modern machine learning methods. In our experiments, we use data on four major currency pairs over a 20-year period. A key contribution is the novel use of outlier detection methods for this purpose. Numerical experiments show that outlier detection methods substantially outperform traditional machine learning and finance techniques. In addition, we show that a recently proposed new outlier detection method PKDE produces the best overall results. Our findings hold across different currency pairs, significance levels, and time horizons indicating the robustness of the proposed method.

Keywords: foreign exchange; forecasting; machine learning; outlier detection; kernel density estimation; KDE; neural networks; tail events.

DOI: 10.1504/IJBIDM.2022.126505

International Journal of Business Intelligence and Data Mining, 2022 Vol.21 No.4, pp.465 - 483

Received: 07 Mar 2021
Accepted: 21 Jun 2021
Published online: 27 Oct 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article