A combination of regression techniques and cuckoo search algorithm for FOREX speculation

This paper describes a hybrid model formed by a mixture of regression techniques and Cuckoo Search algorithm to speculate USD/EUR variations. Inspired by ARMA model we propose a dataset composed of historical data of USD/EUR and (JYN, EUR and BRP) variations. The dataset is used to train four regression algorithms: Multiple linear regression, Support vector regression, Partial Least Squares regression and CRT regression tree; the generated regression weights of these algorithms will be used as inputs to Cuckoo Search algorithm. The effectiveness of the proposed system against classical regression algorithms is confirmed by experiments on exchange rate prediction within the period from January 2014 to January 2016. © Springer International Publishing AG 2017.