Wavelet De-noised Artificial Neural Network (WDANN) for Exchange Rate Predication
Hu, Qiongjie (2010) Wavelet De-noised Artificial Neural Network (WDANN) for Exchange Rate Predication. [Dissertation (University of Nottingham only)] (Unpublished)
This research describes the exchange rate which plays a decisive role in the flow of foreign capital, international banking, international trade and risk management. It briefly reviews the development of exchange rate forecasting models and techniques for financial time series of de-noising wavelet analysis, and then to analyze the applications of the neural network for exchange rate predication. Combined with the advantages of the neural network and wavelet analysis in non-linear forecasting and de-noising respectively, a combinative model wavelet de-noised artificial neural network (WDANN) is developed. This model is tested by empirical researches using GBP/USD exchange rate data. The de-noising performances of the wavelet with different wavelet functions, decomposition levels and wavelet threshold are also tested using experiments respectively. The results indicate that out-of-sample performance is improved by the combinative model. And the following parameters yield more accurate forecasts: coifN wavelet function, decomposing level of three, and sqtwolog threshold.
Actions (Archive Staff Only)