Chemical reaction optimisation: a hybrid technique applied to functional link artificial neural networks with least mean square learning for foreign exchange rates forecasting

Abstract

Forecasting foreign exchange rates has long been an important issue in international finance. Most of the standard econometric methods are unable to produce significant superior forecasts because of its built-in complexity and practical applications. Taking into consideration the worldwide financial capital market, the foreign exchange (FOREX) market has a very crucial role to play. Due to the globalisation of fiscal investment, the investors are interested to learn the co-movement of foreign exchange, so as to make their investments safe and earn profits in return. In this work, an improved ANN model is being proposed that hybridises chemical reaction optimisation with functional link artificial neural network for prediction of foreign exchange (FOREX) rate. Experimental result shows that the proposed model with least mean square (LMS) training outperforms other methods, which ultimately indicates that the proposed model can be an effective way to improve forecasting accuracy achieved by other counterparts.