Using artificial neural network trained with genetic algorithm to model GDP prediction

Using the Gross Domestic Product (GDP)as a measurement of aggregate economic activity is typical. Its cycles are used as an indicator of boom or recession in the economy. Economists and business people are interested in describing and predicting future values of such indicators.However, due to its intrinsic difficulty and non-linear characters, along with many unknown and random events, it is not surprising that forecasts are regularly wrong. The traditional econometrical model cannot give a accurate description and prediction for its linear property. An newly-emerged method for predicting such indicators is to use the Artificial Neural Networks(ANN) trained by Genetic Algorithm(GA). In this paper we establish a new mathematical model to describe and predict the developing trends of GDP based on a hybrid algorithm. We introduce in it the Over-Lapping Generation (OLG) model, thus to set our macro prediction of GDP a micro foundation. Through this blending, we connect the micro agents’ prediction and decision with the macro economic, and therefore endow the new model with economic meanings. Using this model we could make a better description of the macro economical fluctuation and prediction for the GDP developing trends. In the empirical analysis, we examine our new model for its prediction capability and efficiency with a practical example. We use China’s GDP data from 1978-2005, comparing with two other model. In sum, we have established a new method to model the complex connections and variations in macro economy, and introduced a new hybrid algorithm to combine ANN and GA based on the OLG model. This new model has a great importance in econometric study. It could be generalized into other economical prediction, especially the macro economy, and it offers a new way in economical research and at the meantime a new method for the study of ANN and GA.