The Empirical Analysis Of The Expected Rate Of Return Of China Stock Markets Based On The Artificial Neural Net Technique

This thesis uses the Artificial Neural Net technique to empirically analyze the expected rate of return of China stock markets.For nearly 30 years, lots of scholars have done much work to test the validity of CAPM theory, in which B is considered to be able to fully explain the stock returns. The scholars mainly used Fama-MacBeth cross-sectional regression model. The empirical results largely hold for the facts that 8 has no explanatory power for the average stock returns. Otherwise, another basic variable, systematical risk of stock, is proved to contribute to the expected rate of stock returns.Basing on the research of domestic and foreign scholars, this thesis sets up a 5-factor model of stock returns on China stock markers. Besides the basic variables (B, 2,size, and E/P), trading volume has been adopted as a new independent from CSMAR database by using Visual FoxPro programming. ANN tools in MATLANB software are used to extend the empirical study on the relations between the basic variables and the stock returns. A 4-factor model is also set up to analyze the stock returns further.ANN technique and F-M technique are compared in the thesis. The results of ANN show that, the variable of size has the absolutely strong explanatory power, while E/P absolutely weak. Meanwhile, trading volume, B and 2 have some more, ANN has three main advantages over the F-M technique. First of all, ANN can variables, while F-M can only reflect the linear relationship. Secondly, ANN can simulate the function for the input and output variables in the model, but F-M can only give some statistics parameters. Thirdly, ANN provides excellent extension capability. In the thesis, a perturbation post-processing method is adopted to calculate the influence factor for the relative change of stockreturns.