Contents

                                              Forex Forecasting

The Backward Propagation Algorithm is presently used by commercial software in the training of neural networks for prediction of forex rates. An example of such a software is NeuNetPro. Though various factors influence forex rates, a statistical report (Waghela – Mallya) obtained a reasonable amount of accuracy in forecasting using only the following factors:
1. Call money rates
2. 91 Day – Treasury bill yield
3. WPI
4. GM LIBOR
5. Reserves
6. US Inflation

Conclusion:

Due to the ability of neural networks to generalize very well, they can be employed wherever something varies as a very complicated function of a number of variables, and where it is difficult to define the function from the data we have. For instance neural networks may be used by direct mail advertisers to decide out of a database which customers to target. More complex uses include handwriting recognition, fingerprint matching, facial recognition and so on.

Bibliography

1. Neural Networks, A comprehensive foundation – Simon Haykin
2. Forex Forecasting – Prashant Waghela and Shailesh Mallya
3. http://www.cormactech.com/neunet/
4. comp.ai.neural-nets FAQ
5. A case study on using neural networks to perform technical forecasting of forex, Neurocomputing, Volume 34, Issues 1-4, September 2000, Pages 79-98 Jingtao Yao and Chew Lim Tan
Copyright IITBombay 2005