Development and performance evaluation of a novel knowledge guided artificial neural network (KGANN) model for exchange rate prediction

This paper presents a new adaptive forecasting model using a knowledge guided artificial neural network (KGANN) structure for efficient prediction of exchange rate. The new structure has two parallel systems. The first system is a least mean square (LMS) trained adaptive linear combiner, whereas the second system employs an adaptive FLANN model to supplement the knowledge base with an objective to improve its performance value. The output of a trained LMS model is added to an adaptive FLANN model to provide a more accurate exchange rate compared to that predicted by either a simple LMS or a FLANN model. This finding has been demonstrated through an exhausting computer simulation study and using real life data. Thus the proposed KGANN is an efficient forecasting model for exchange rate prediction.