Economic Early-Warning Based On Rough Set And Constructive Learning Neural Networks

As Economic early warning is the cognition and judgment of economic status, its results affect the choice of macroeconomic policies, it has been one of the most important research fields of economics, and widely concerned about by all governments and public. However, traditional economic warning methods, which are based on experts' experience or simple statistical models, cannot reflect the nonlinear nature of economic system. So it is hard to deal with nonlinear problems thereby, it cannot meet the demand of macroeconomic early warning. This paper firstly discusses the recent development of macroeconomic early-warning and points out the present problems in macroeconomic early-warning indicator system designing, early warning models and the early-warning system establishing, after analyzing the theory and method of artificial neural network and rough set, the feasibility of early warning combining the rough set with constructive machine learning is discussed and the algorithm of macroeconomic early warning system based on rough set and constructive learning ANN is put forward. The experiment using the monthly data of Anhui province shows that the model is feasible and high efficiency. The research work supplies a new way for dynamic economic early warning.Main work and results include:(1) The selecting of economic early warning index system based on rough set. Since there exist correlation and redundance between economic early-warning indices, after introducing the theory and approaches of rough set, the principle, strategy and algorithm for the selecting of economic early warning indices based on rough set is presented.(2)Concretely addresses the model of economic early warning based on rough set and covering-algorithm, which include the thought, feasibility analysis, concret designing and algorithm of the model.(3)The economic early-warning model is testified through data experiments in practice and compared with BP.