Application of Neural Networks for Foreign Exchange Rates Forecasting with Noise Reduction

Abstract

Predictive models are generally fitted directly from the original noisy data. It is well known that noise can seriously limit the prediction performance on time series. In this study, we apply the nonlinear noise reduction methods to the problem of foreign exchange rates forecasting with neural networks (NNs). The experiment results show that the nonlinear noise reduction methods can improve the prediction performance of NNs. Based on the modified Diebold-Mariano test, the improvement is not statistically significant in most cases. We may need more effective nonlinear noise reduction methods to improve prediction performance further. On the other hand, it indicates that NNs are particularly well appropriate to find underlying relationship in the environment characterized by complex, noisy, irrelevant or partial information. We also find that the nonlinear noise reduction methods work more effectively when the foreign exchange rates are more volatile.KeywordsPrediction PerformanceHide NodeNoisy DataRandom Walk ModelLocally ProjectiveThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.