The Research On Forecasting Of Stock Price Based On Wavelet Neural Networks And STAR

All along, the scholars have not stopped the nature of the stock market trend forecasting and market research. Formed during this period many of the theories and research results. Its stock index as an important financial data, the uncertainty has a strong and nonlinear, which makes the study and forecast the stock index is very difficult. In the current study, intelligent algorithms based on neural network comparison of the effects achieved outstanding as on the stock market one of the main research and extension on this basis to develop different models. Through the study found that the stock market, stock market trend with a long-term and short-term uncertainty of two characteristics, the two characteristics can be studied separated from each other more accurate results. Wavelet analysis has a very good time-frequency localization properties, being used more and more to the economic and financial fields, and then for time series data to achieve good results. At the same time the stock market in the short-term fluctuations in the characteristics shown are the mean reversion in this study were taken into account, and through the STAR model study.On this basis the advantages of various research methods will be the integration, decomposition of the characteristics of the index to describe and explain, and then reconstructed by wavelet new forecasting model. Focus of the study is based on wavelet neural network and the advantages of STAR models to build new forecasting model, and application and implementation. Main tasks are as follows:First, the background and research on topics of significance were explained. Then describes the current stock index time series for a major research and forecasting methods and problems, and discusses the characteristics of the mean reversion index, form and test methods.In Chapter II, the characteristics of Chinese stock market volatility and its influence factors were studied, obtained the asymmetry of the market volatility characteristics. Proposed prediction model for further lay the foundation for the stock market. Chapter III describes the wavelet-based neural network and STAR (WNN-STAR) model theory, introduced the STAR model and the basic theory of wavelet neural network and applied research, described in detail the theoretical basis of prediction models and modeling basis. Chapter established based on wavelet neural network and STAR of the forecasting model of stock index, stock index by using the wavelet analysis will be divided into two parts, low and high frequencies, respectively, the neural network fitting the trend of low frequency data, high frequency data with the end of STAR models The mean recovery, and then merge the results by wavelet reconstruction, resulting in more accurate and scientific explanation and prediction. Chapter V of the Shanghai index by selecting the empirical data and the results were analyzed and compared the results have been more satisfactory, further evidence of the present model of scientific and practical.