| Economic globalization is not only an opportunity for China to develop great-leap-forward, but also to bring us more challenges. The main problem we must consider is how to accurately grasp the short-term and long-term trends in economic development and serve the rapid economic development. It is very difficult to obtain good simulation results using the traditional economic forecasting methods, such as certain time series analysis method(Exponential Smoothing, Moving average, time alignment decomposition ) ,Mutuality (Regression) analysis, Gray forecasting theory and combinatory forecasting model. And using traditional forecasting methods, there exist many difficulties such as multi-collinearity, error sequence related etc, and the foresting precisions are not satisfied.GDP forecast is a difficult problem because of its dimensions of input vector, the impact of complicated factors, highly nonlinear, the coupling complex components of the input vector, being a typical "black box" model and without clearly expression between the direct interactions of data. In this paper we consider that GDP can be attributed to many factors, but only seven plays a major role: The Country's GDP, Total Export Value of Foreign Trade, Local Government Expenditure, Total Retail Sales of Consumer Goods, Value of Foreign Capital Actually Used, Local Government Revenue and Investment in Fixed Assets.Artificial neural network is a nonlinear, non-local, non-stationary complex network system. It has parallel distribution and adaptive structure of the brain as an information processing model. And it can complete specific tasks by "self-learning" or "training" to learn a large numbers of knowledge. Economic forecasting modelusing artificial neural network has high accuracy.Many practical examples show that the artificial neural network is very effective method to construct forecasting model. It can learn knowledge from the sample data without complicated inquiry and expression; automatically obtain the functions behind the sample data. It is superior to the traditional time series forecasting methods obviously, especially when the model has strong complexity nonlinear. The forward neural network is most widely used and successful model for forecast. BP networks, which often used to make prediction, has simple structure. It can solve the multi-factor complex nonlinear problems which traditional forecasting techniques can not solve.Comparing time series analysis model with econometric measure model, it is simpler and more operational. The forecasting accuracy base on neural network is better than that of vector auto-regression model.This paper studies the GDP forecast using the theory of artificial neural networks. First, we give some introductions for the domestic GDP forecast methods and artificial neural networks. Then, we employ a three feed-forward back-propagation neural network to construct forecasting model for GDP forecast using the GDP data of Guangxi. When training the networks, we normalize input and output data for the same magnitude. And the predictions are compared with the traditional model; the results show that the GDP forecast using BP model is more available. |
