Research On Security Market Forecasting Method

The development of our country's stock market is fast. The forecasting methods of security market are various. The method of quantitative analysis is consists of the linear and nonlinear method. The linear method be used to forecast the stock market price which often has some localization. Now the nonlinear method is more and more abundant, which has been show its advantage in stock market.This thesis simply introduces the linear and nonlinear method and their present conditions, expounding the research status of the neural network and the wavelet neural network. Then, we set up the linear time series model to forecast the closing price of the ChangJiang electric power.The neural network has mighty nonlinear ability. It was been widely used in many fields. The thesis uses the Back-Propagation method, which is frequent of the neural network,building the three layer network for the forecasting of the closing price of ChangJiang electric power. Finally, according to closing price curve, the wavelet neural network was set up to forecast it. In the process of designing the wavelet neural network, the thesis discusses the way to initialize the parameter of the network and uses the self-realization step method to the study process of wavelet neural network.With the comparison of these three kinds of forecasting methods, the thesis makes a sum-up. In these forecasting methods, the nonlinear method often can get a better forecasting accuracy than the linear method. The wavelet neural network has a characteristic of self-training. It has faster convergent speed and higher precision than the traditional neural network. It has good foreground in the nonlinear forecasting fields.