Forex Daytrading Strategy : An Application of the Gaussian Mixture Model to Marginalized Currency pairs in Africa

Abstract

In this article, we build a Daytrading strategy by applying a Gaussian Mixture Model (GMM) to so-called marginalized African currency pairs. With a sample of data covering the period of 01/01/2010 to 05/15/2021 and downloaded in real time from the US Federal Reserve (FRED) website and the Yahoo Finance platform, we found that by including four explanatory variables in the Gaussian mixture model, the GMM independently estimates the returns of the following six African currency pairs with an average accuracy of approximately 56.379%: USDZAR, USDNGN, USDEGP, USDMAD, USDMUR and USDKES. The accuracy of the estimations obtained with the GMM on the returns of the USDZAR currency pair is the highest (73.887%) among the six currency pairs studied in this work. Based on the above results, we have built a trading robot (based on the GMM) that runs in real time and which we have deployed in production in the exchange rate market. In general, by adjusting several parameters of the GMM, our trading robot achieves overall positive daily gains.