Foreign exchange prediction and trading using deep belief neural network

Abstract

Foreign exchange rate prediction is never an easy task due to the size of the foreign exchange market and the different influences in the market. In this report, neural network is used as a modelling technique to predict currency prices. Continuous Restricted Boltzmann Machine (CRBM) makes use of a training algorithm to model continuous data, which is the building element of the model. By stacking the CRBMs, the Deep Belief Network is created to forecast one-step ahead predictions. Experiments are performed to evaluate the effects of weight updating methods of CRBM and measured in terms of Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The results suggest that deep belief network should be complemented with other analysis tools to make better trading decisions.