Study of Canada/US Dollar Exchange Rate Movements Using Recurrent Neural Network Model of FX-Market

摘要

Understanding exchange rate movements has long been an extremely challenging and important task. Unsatisfactory results produced by time series regression models have led to the claim by several authors that in foreign exchange markets, past movements of the price of a given currency have no predictive power in forecasting future movements of the currency price. In this paper, we build a recurrent neural network model for FX-market to explain exchange rate movements. Asset prices are discovered in the marketplace by the interaction of market design and agents' behaviour. The interaction is simulated by integrating 1) the PX-market mechanism; 2) an economic framework; and 3) the embedding of both tasks in neural network architectures. The results indicate that both macroeconomic and microeconomic variables are useful to forecast exchange rate changes. Results from regression model based on neural-fuzzy forecasting system are also included for comparison.