The Application Of Data Mining In Yield Volatility Forecast Of Shanghai And Shenzhen 300 Index

In the last few decades, Security market grew up quickly in China. More and more people put their money into the security markets which flourished recent years, in order to gain a great of return. At the same years, Data Mining technology is looked forward to dig potential information from magnanimity financial data with the rapid developing of Securities market, so researching on data mining area has become an important issue in financial analyzing area.From established in 2005, Shanghai and Shenzhen 300 Index has become an indication that the Shanghai and Shenzhen stock market characteristic. The selection criteria are large-scale, good mobility. The index covers more than 60 percents market value in Shanghai and Shenzhen stock markets. It is the underlying asset of upcoming stock index future too. The yield of the index forecast can be viewed from two perspectives: One from the source of yield. The impact on the yield of stocks from the source of the external environment, regulatory authorities, the fundamentals of listed companies, investors, technical analysis and related investment goods prices . Another from fluctuations in the time sequence, the stock yields a nonlinear time series addition, the general non-stationary time series with the characteristics, it also has thick tail peak, high noise, such as aggregation of volatility characteristics.For the above two perspectives analysis approach, we use data mining methods for analysis and forecasting. First, from the volatility of the six sources to find some representative indicators, the paper use data mining in the Logistic, decision tree and neural network method to analysis and forecast the indexes Change trend. Find the source of the volatility of some of the features, and found that neural networks in predicting the results were very good. BP neural network is then used in the adaptive, self-learning, non-linear optimization and characterization of GARCH model of the time series of auto-correlation, aggregation of volatility, the peak thick tail. The establishment of the BP-AR-GARCH model forecast and has achieved good results.