Predicting exchange rates by support vector regression

In recent years, support vector regression (SVR), a novel neural network technique, has been successfully used for financial forecasting. This paper deals with the application of SVR in exchange rate forecasting. Based on SVR, a nonparametric autoregressive (AR) model is applied to forecasting the daily exchange rates of two currencies (South Korea Won and Singapore Dollar) against the US dollar. The empirical results show that under various forecasting horizons, SVR performs better than the random walk model, parametric AR model and the nonparametric AR model estimated by neural network, based on the criteria of two evaluation metrics and three encompassing tests. No structured way being available to choose the free parameters of SVR, the sensitivity of the forecasting performance is also examined to the free parameters.