Loading...

Modelling and Forecasting Exchange Rates via Econometrics Models and Neural Networks

Sofiazizi, Aziz | 2015

615 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 47386 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Kianfar, Farhad
  7. Abstract:
  8. Due to the significance of exchange rates in economic policy making, different patterns have been proposed so as to explain the behavior, provide ways to model and deliver tools to forecast different exchange rates. Using a novel approach, this thesis tries to investigate the behavior of exchange rates by identifying time series nature of exchange rates, and performing nonlinear test for daily data between years 2003 to2006. In this study, we try to model and forecast the daily exchange rates by the use of Artificial Neural Networks (ANN). We also compare the results with ARIMA model outputs based on measures for forecasting accuracy. 80 percent of the daily data, that is, 1160 days from March25, 2003to June5, 2006 has been used to train the models. In order to perform the sensitivity analysis, data for Dollar, Euro and Pound have been considered. The results show the better forecasting ability for the Neural Networks compare to ARIMA model. Besides, for a certain day, exchange rates for Euro and Pound are functions of data from previous day of Euro and Pound respectively, and Dollar exchange rate is a function of data from previous 6 days of Dollar exchange rate. What distinguishes this study from others is the unique design of the artificial neural network which by using an activation function and a learning algorithm can minimize the prediction error, and approximate any arbitrary function with any degree of accuracy that is needed
  9. Keywords:
  10. Exchange Rate ; Forecasting ; Artificial Neural Network ; Autoregressive Integrated Moving Average (ARIMA) ; Econometrics ; Rate

 Digital Object List

 Bookmark

No TOC