Shanghai Composite Index Return Predictability Studies

With the rapid development of the Card coupons market and the continuous improvement of the income level of resident, and at present there are more and more people involved in the stock market investment, hope to be able to realize the asset value maintained and added value. The stock market while as a possible higher returns to investors with the investment direction, but at the same time it with higher risk。. Its market price changes of uncertainty, let many investors feel the complexity of this market, so they are in desperate need of a theory of price change analysis of the causes, and needs to be a scientific forecasting method to guide the stock investment, so as to realize the investment risk, avoid a higher return on investment objectives. Stock price forecast always is a more difficult task, but at the same time the study if there is some breakthrough, also contains the enormous economic interests.Peters in the early1990s put forward the fractal market hypothesis, think of all the stable market in fractal structure, stock prices to a certain extent existing prediction. The behaviors of investors are influenced by the stock market information understood and investment the length of time.In acknowledging the stock price exist under the premise of predictability, this paper briefly introduced the stock price prediction methods, which include not only the traditional basic analysis and technical analysis, but also including recent popular mathematical statistical prediction method.This paper focuses on the research of the Shanghai composite index excess return of predictability:A Bayesian Model Selection Perspective.This paper Comparative analysis the Prediction ability among overall posterior probability weighted average models for various pand for both the case of weak and information priors, the "best" individual models according to the statistical model selection criteria, the individual models with highest posterior probability for various ρ and the constant, unconditional model, in-sample performance. And then in the sample is whether there is also good consistency performance. Conclusion, the use of bayesian method in out-of-sample performance there are still good prediction ability.