Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/238807 
Year of Publication: 
2013
Citation: 
[Journal:] International Econometric Review (IER) [ISSN:] 1308-8815 [Volume:] 5 [Issue:] 2 [Publisher:] Econometric Research Association (ERA) [Place:] Ankara [Year:] 2013 [Pages:] 53-69
Publisher: 
Econometric Research Association (ERA), Ankara
Abstract: 
ANN (Artificial Neural Network) models and Spline techniques have been applied to economic analysis, to handle economic problems, evaluate portfolio risk and stock performance, and to forecast stock exchange rates and gold prices. These techniques are improving nowadays and continue to serve as powerful predictive tools. In this study, we compare the performance of ANN models and Bayesian Spline models in forecasting economic datasets. We consider the most commonly used ANN models, which are Generalized Regression Neural Networks (GRNN), Multilayer Perceptron (MLP), and Radial Basis Function Neural Networks (RBFNN). We compare these models using BayesX and Statistica software with three important economic datasets: on the exchange rate of Turkish Liras (TL) to Euro, exchange rate of Turkish Liras (TL) to United States Dollars (USD), and Gold Price for Turkey. With these three economic datasets, we made a comparative study of these models, using the criterions MSE and MAPE to evaluate their forecasting performance. The results demonstrate that the penalized spline model performed best amongst the spline techniques and their Bayesian versions. Amongst the ANN models, the MLP model obtained the best performance criterion results.
Subjects: 
Artificial Neural Networks
Bayesian Spline Models
Exchange Rates
JEL: 
C11
C45
C53
Creative Commons License: 
cc-by-nc-nd Logo
Document Type: 
Article

Files in This Item:
File
Size





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.