Recurrent Self-Organising Maps and Local Support Vector Machine Models for Exchange Rate Prediction
This paper considers the problem of predicting non-linear, non-stationary financial time sequence data, which is often difficult for traditional regressive models. The Self-Organising Map (SOM) is a vector quantisation method that represents statistical data sets in a topology preserving fashion. The method, which uses the Recurrent Self-Organising Map(RSOM) to partition the original data space into several disjointed regions and then uses Support Vector Machines (SVMs) to make the prediction as a regression method. It is model free and does not require a prior knowledge of the data. Experiments show that the method can make certain degree of profits and outperforms the GARCH method.
KeywordsSupport Vector Machine Local Model GARCH Model Time Series Forecast Time Series Prediction
Unable to display preview. Download preview PDF.
- 5.Vesanto, J.: Using The SOM and Local Models in Time-Series Prediction. Helsinki University of Technology B15 (1997)Google Scholar
- 7.Koskela, T.: Time Series Prediction Using Recurrent SOM with Local Linear Models. Helsinki University of Technology (2001)Google Scholar
- 9.Kohonen, T.: Self-Organizing Maps. Springer, Berlin (1995)Google Scholar
- 11.Walter, J., Ritter, H., Schulten, K.: Nonlinear Prediction with Self-Organizing Maps. Beckman-Institute and Department of Physics, UIUC, UrbanaGoogle Scholar
- 14.Koskela, T., Varsta, M., Heikkonen, J., Kaski, K.: Time series Prediciton Using Recurrent SOM with Local Linear Model. Report B15, Lab. of Computational Engineering, Helsinki University of Technology (October 1997)Google Scholar
- 15.Yin, H.: Self-Organising Map as A Natural Kernel Method. In: Proc. ICNN&B 2005, Beijing, China, October 13-15 (2005)Google Scholar
- 16.Kuhn, H.W., Tucker, A.W.: Nonlinear Programming. In: Proc. 2nd Berkeley Symposium on Mathematical Statistics and Probabilistics, vol. 481C492, University of California Press, Berkeley (1951)Google Scholar