Prediction Of Shanghai Stock Exchange Composite Index Based On RBF Neural Network

The stock market is an important part of Chinese capital market, and it plays an important role in prompting the economic development in China. For better understanding and getting more gains, forecasting has gained more and more attention in the last decade. On the stock market, the SSECI(Shanghai Stock Exchange Composite Index)is one of the principal financial indicators to make judgment, that is why forecast the trend of SSECI is realistically and valuably.At least 15 models have been proved to be effective in forecasting, RBFNN is one of them. Because of its simple structure RBF neural network had been used as one of the hot models in recent years. It forms a special architecture of neural network, which has advantages of the simplicity of its structure, faster learning algorithms, and better approximation capabilities.This paper first introduced the development of the stock market in China and some knowledge of stock forecasting, then gave some basement of neural network theory. After that the paper examined the applicability of RBFNN(Radial Basis Function Neural Network) used to forecast the SSECI .Using the data of closing prices from 2006 to 2009, the paper first divided the data into four periods, in each of which had the same sets of data used for training and testing. Then experiments have been designed and simulated by Matlab7.0 with the purpose of testing the applicability of RBFNN used to forecast the SSECI in each particular period and the purpose of finding the proper size of data sets for training.. It was shown that obviously differences were in each period and classified the date before using RBFNN to forecast was needed and 140 was the first proper size of the data for training, and RBFNN should be used in medium-short term prediction longer than a month while shorter than 8 months, especially in the prediction of SSECI.At last, the paper compared the results of final one-step prediction and the results of rolling forecasting, and then forecasted one period of the Shanghai Stock Exchange Composite Index...