چکیده:
Applying nonlinear models to estimation and forecasting economic models are now
becoming more common, thanks to advances in computing technology. Artificial
Neural Networks (ANN) models, which are nonlinear local optimizer models, have
proven successful in forecasting economic variables. Most ANN models applied in
Economics use the gradient descent method as their learning algorithm. However,
the performance of the ANN models can still be improved by using more flexible
and general learning algorithm. In this paper, we develop an ANN model combined
with Genetic Algorithm to forecast the Iranian GDP growth. In order to evaluate
the performance of the model with other ANN and traditional econometric models,
we compare the results of the model with other linear and nonlinear competing
models such as ARMA, VAR, and ANN with gradient descent learning algorithm. We
use the recently produced extended data on the Iranian GDP from 1937 to 2002.
The results indicate that the GA can improve the forecasting performance of ANN
model over other standard ANN and econometrics models.
خلاصه ماشینی:
"Keywords: Forecasting, GDP Growth, Artificial Neural Networks, Genetic Algorithm, ARMA, VAR 1- Introduction Applying nonlinear models to estimation and forecasting economic models are now becoming more common, thanks to advances in computing technology.
4- ANN Model A set of ANN models with back-propagation and gradient descent learning algorithm are applied to estimate and forecast the Iranian economic growth.
Although the ANN models can learn any patterns in the data, as Swanson & White (1997), Chatfield (1993), and Moshiri & Cameron (2000) show, the best results are produced when the input variables are selected based on the economic theory or the statistical features of the data.
In this paper, we apply the test developed by Diebold-Mariano (1995) and its revised version by Harvey, Leybourne, and Newbold (1997), to check for statistically meaningful discrepancies among the forecast errors generated by the GA, ANN and other econometric models.
According to RMSE and MAE criteria, ANN with different specifications have Table (4): Forecast Error Comparisons among Competing Models for the First Two Years of the 3rd Iranian Economic Plan 2001 2002 RMSE MAE Actual Growth 5.
The GA, which is rooted in biology, searches for the best-fit solutions using crossover, mutation, and selection operations, and is particularly useful when the search space of possible solutions to a given problem is high dimensional, where the systematic search algorithm is not feasible due to sheer number of combinations to test (McNelis, 1997)."