Research Of Gdp Forecast Based On Self-organising Data Mining

GDP forecast has been the focus of the scholars. Contraposing the problems and shortcomings of the modeling in the past, in order to avoiding the factitious effect, retaining objectivity and veracity of the model, a new method is suggested to study GDP forecast------Group Method of Data Handling (GMDH).GMDH is a method of designing model by data handling, which is similar to the process of evolvement, and it make easier to understand the discipline by the result of model. GMDH applied mathematics combination method and self-choice method which based on different inspired standards, meet the objective and the authenticity of the requirements.This paper use the method of self-organizing data mining to forecast GDP based on macro economic data. The main content is algorithm improvement and realization the algorithm. In the part of improvement, the paper puts forward the improved GMDH algorithm------multiple GMDH method, which not only to retain the original variables, the more strict in parameter setting, and set the new effective stop law. In the part of algorithm realization, the paper proposed several new method of data pretreatment, and used the improved method in real GDP forecast model in China. The model reserves four effective variables, and explains the results more reasonable. At the same time, the model precision is satisfactory. Finally the paper compares the new model with other models, such as neural network, time series, MLR (Multiple Linear Regression), and finds the new method is superior to all the three methods.