Research & Application Of BP-RBF Combination Neural Network Forecasting Method On Stock

With the development of the national economy and the perfection of market, the stock market has gradually become an integral part of the China's securities and the whole financial industry. However, the development of China's stock market has a short time, still immature. Irrational investment still exist in large numbers. Effectively predict the stock market has great theoretical and significance. In recent years, scholars through the multiple regression analysis,time series analysis, exponential smoothing methods to predict, however, due to non-linear characteristics of the stock market, traditional methods is difficult to give a more satistactory results.Neural network as an effective intelligent information processing technology, has achieved remarkable development and has been widely used. It has learning, memory and computing functiongs, has become a power tool to solve the problem, to break through the bottleneck of current science and technology, more in-depth exploration of the complex nonlinear phenomena played a major role, and is widely applied in many fields of science.In this paper, BP,RBF neural network and its combined models are used to predict the stock. We first deeply investigate and research the basic theory of netwoks; then use the BP and RBF neural network to establish a single short-term forecasting model, structural design,the parameters of the model selection, data preprocessing and other issues were discussed. And we conduct an empirical analysis as a example of Chinese Banks. In order to further improve forecast accuracy and stability, this paper presents BP and RBF neural network based on master-slave neural network model and the fitting error combined neural network model and conduct an empirical analysis, results showed that:Composite neural network model achieved desirable results both in accuracy and stability than single neural network model. So it will be widely used in the future.