Nonlinear Modeling And Forecasting Of Money Demand Function Based On Generalized Regression Neural Network

Money supply is the intermediate target of monetary policy. The adjustment of money supply will have great effect on the final target of the monetary policy. For taking hold of the money supply, we need to analyze money demand from the angle of the balance of money market. So, a stable, reasonable and forecastable money demand function is very important for effective money policy.The macroeconomic system is a very complicated system, with extensively nonlinear, time-varying and the indetermination function relations. Especially at present, our country is at the stage of socialist market economy system awaiting reform and perfection. Some factors are very difficult to reflect in the linear model, even can be reflected under some situations, explain extent is not enough. In this case, it forces people to seek a kind of non-linear tool and carry on economic modeling. The highly approaching nonlinearity ability of the artificial neural network has offered a brand-new way for macroeconomic analysis. Given the traditional BP neural net has problems that the convergence speed is slowly and the net maybe trapped into local minimum, it is limited in applications. So we need another artificial neural net to built the money demand function. Generalized Regression Neural Network is very effective because of the ability of nonlinear mapping and the property of fewer samples for modeling than BP net.Compared with training process of the BP net and GRNN, this thesis gave the choice of the neural net in solving nonlinear model of money demand. The macroeconomic quarter data of variables during 1996 to 2006 was choused to estimate the nonlinear error correction model and build the nonlinear money demand function based on GRNN. Compared with the forecasted results of cubic polynomial error correction model for Ml and M2, the GRNN model is effective and Anti-Interference. We used this model on the application of forecasting the money demand of 2007.And it is proved feasibility compared with the real dates.