Research On The Prediction Of Shanghai Composite Index Based On Neural Network

The stock market as a social economic development plays an important role in financing and resource allocation, its relatively stable earnings and moderate risk attracts more and more investors to participate in them, how to effectively analyze and forecast the trend of the stock market has become a problem of concern to the people, the traditional analysis method has been adopted widely, but the traditional method cannot avoid the Human intervention, because of its special structure and methods of information processing, highly parallel, distributed storage and so on, the artificial neural network is particularly suitable for dealing with uncertain fuzzy information and problem needs to consider many factors conditions at the same time,that provides an effective method for a quantitative the analysis for prediction of stock market.However, for complex issues and high-dimensional input variables, the direct use of neural network prediction, will bring dramatic increase in network size, increase in computing time, and reduce the network convergence and generalization ability; on the other hand, due to correlation between predictors, leading to input information overlap, but also makes the accuracy of the model prediction to reduce the and, therefore, the necessary processing to deal with such samples and numerous forecast factor.According to the forecasting closing price in stock market, in order to improve the efficiency of the prediction accuracy and training network, many of the forecast factors by principal component analysis to reduce the dimension of the treatment, we can use SOM neural network to classify for large sample, for each sub class after the classification, respectively, to establish the BP neural network and the corresponding prediction model; in actual forecasting, we must classify real-time data, select the corresponding model to predict the output. Thus, the BP neural network, principal component analysis, SOM neural network organically, to construct the real-time combination of closing stock price prediction model. Through the empirical study proves that the improved BP neural network model to the effectiveness of stock price prediction.