Paper
An Application of Neural Network and Hidden Markov Model: A Case Study for Euro Dollars Exchange Rate
- Authors:
- Majid Rastegar; Saeed Rasekhi
- Abstract
- In this paper a robust hybrid algorithm for forecasting exchange rate trends is proposed and simulated. The existence of noise in time series of exchange rate causes many forecasting methods encounter with uncertainty prediction degree. This work introduces a coefficient framework based on artificial neural networks (ANNs) tools to overcome this problem. First time series samples of exchange rate are quantized by self organizing map (SOM) which is considered a powerful tool in ANNs. SOMs are different from other artificial neural networks (ANNs) since they use a neighborhood function that has the ability in preserving the topological of the input space. A signal processing at this level can be considered as a signal smoothing operation for denoising input data. In the next level the proposed algorithm uses Elman neural network for estimating the direction of quantized (training/test) input data. The simulated result shows that application of SOM neural network can overcome some limitations of conventional methods. To show the effectiveness and efficiency of proposed algorithm, the performance of the model is compared with a Markov model by using daily data of 7 November 2002 to 9 October 2006 to predict the trend of exchange of Dollar/Euro. The simulated results show that Markov model has superior predicting performance than proposed hybrid algorithms.
- Keywords
- Exchange Rate; ANN; Markov Model; SOM Neural Network; Dollar/Euro
- StartPage
- 169
- EndPage
- 175
- Doi