| The stock market is one of the most important parts of the financial market. The trend and fluctuation of stock price can reflect the politics,economy and social situations of a country and conduct macro control for Countries. However, the stock market is a very complicated and nonlinear dynamic system, there are many factors affecting the changes of stock price. It is not easy for investors to analyze the trend of the stock price because it is usually of ups and downs.Then,does the Chinese stock market exist the mean reversion phenomenon? Is the trend of stock price predictable?In this paper, based on the research of the existence of mean reversion in China stock market,we analyzed the prediction techniques of stock market and proposed a combined predictive model. The paper are divided into two parts as follows:In the first section, this paper tests the mean reversion of the monthly industrial index, the commercial index,the real estate index,the utility index and the composite index in Shanghai Security Exchange from January 1996 to June 2015 with the auto correlation test and cointegration verification. Empirical results show evidence of mean reversion in China stock market,which isn’t efficient enough to follow a unit root process. Therefore, the future trend of the stock price can be predicted.In the second section, this paper introduces the basic principle, modeling process of the SARIMA model and BP neural network in detail. The SARIMA model is a generalization of the ARIMA model in seasonal time series, which has a strong capability on linear modeling.Compared with other models, the BP neural network model is a nonlinear model,which has incomparable advantage on the forecasting of the nonlinear time series. In this paper, we propose a combined predictive model SARIMA-BP model and use them to fit and forecast the Shanghai composite index series. The experimental results show that the combined model can be applied to the prediction of stock price time series. The precision of prediction is more accurate than the result of single SARIMA model or BP neural network. |
