Prediction On Stock Index Based On Grey Neural Network

In the modern economy, the stock index plays an increasingly important role, As we know, if one can grasp the development trend of the stock and make the right decision in a shorter time than others, it helps the one to get more benefits, and be a winner in the stock market, the research work of the stock index has been put more and more attention by analysts and investors all over the world. With the development of computer technology, the Grey Neural Network model was began to apply in forecasting the stock index, and the results proved the model could get a better predictive accuracy, what's more, the model needs less samples and faster trainning time, the model provides a new predictive method for stock index forecast.First, the paper outlines the research of grey system and neural network technology, on the basis of these, the paper studies the algorithm of the two single forecast models deeply- BP and GM(1,1), the shortcomings of the BP algorithm are found and improvement methods are also proposed behind the detailed analysis, then, a new BP network was designed in this paper. At the same time, the paper also focuses on GM(1,1), which was one of the core predictive model of the grey system, the GM(1,1) application scope is discussed and obtained relevant conclusions in this pepar.The paper studies emphatically on five types of forecast model, including BP, GM(1, 1), GM(1, N), GNNM(1,1) and GNNM(1, N). The five models separately introduce the the algorithm and creating process of the model, in this five models, the first three models are single forecast model, and the last two models designed solution methods for one-dimensions and N-dimensions grey problems.The Shanghai Composite Index closing price is considered as the research object of this paper, through analysising of the grey relational degree of 11 most used factors with stock index, choosing out two factors with a bigger grey relational degree to be the inputs of the grey neural network, then, from the large samples and small samples, the paper designs three prediction models to study the predictive accuracy of grey neural network model, the three models including BP neural network, GM(1,1), GNNM(1, N). The examples shows GNNM(1, N) has a better prediction accuracy then BP neural network and GM(1,1) in the small samples. Finally, Microsoft Visual Basic 6.0 associated with MATLAB 7.0 is used together to develop stock index predictive programe. The programe offers three models for user to select, including BP neural network, GM(1,1), and GNNM(1, N).