Macro-Economic Early-Warning Wethod Reaseach-Kohonen-Bp Neural Network Model

In this paper, on the domestic and international macro-economic overview of early-warning method based on the various methods will be summed up as traditional indicators of early-warning method, the traditional boom lights law, econometric models of early warning early warning method and pattern recognition method of four broad categories, and from the basic principles, excellent poor nature of the angle of the applicability of this method of theoretical study four methods were put forward ideas for possible improvements. The study is to further deepen the macroeconomic early warning method for people to correctly understand and use scientific and rational ways to provide a reference basis for the theoretical model of innovation made after the bedding.This warning followed the field of neural networks used in the macroeconomic and development of new hot spots, in-depth analysis of the Kohonen and BP neural network model topology, algorithm theory, model features and applications of the difference between the two models were compared with analysis, on this basis proposes a Kohonen-BP neural network based on early-warning model and the structure of the model building process.. The basic idea of the model:the use of Kohonen clustering analysis of data, and the results obtained with experts warning given degree of category analysis and revision, so that the revised sample data input BP neural network training, the final use of neural network output to conduct early warning analysis. Its main purpose is to further deepen the level of macro-economic objective of early warning and strengthen its handling of complex and highly non-linear economic early-warning capabilities.Finally selected industrial cash expenditures, total retail sales of social consumption, industrial output, electricity, the budget of industrial sales revenue, the total investment in infrastructure, currency in circulation, the customs value of total exports and mentor to the nine indicators of the data model was empirical study to verify results show that the model has good predictive power. The entire process of using SPSS Clementine12.0 operation.