The application research of MG model on the simulation of financial markets in China

Challet, Marsili and Zhang built a ''Producer-Speculator'' financial market with MG model, and thus regenerated the stylized facts of financial markets. We modified the model by including noise-traders to simulate the real financial market of China in an improved way. Autocorrelation function analysis and R/S analysis (rescaled range analysis) was employed in the paper to test the assumptions on the "clustering" phenomenon and ''fat-tailed" distribution of the return rate of financial markets. The research showed that the financial markets in China couldn't be regarded as efficient markets on a frequent observation time-scale (with the frequency as high as one observation per 15 minutes). The Hurst exponent is bigger than 0.5 and the autocorrelation is distributed between -0.05 and 0.5. Moreover, by including noise traders, the behavior of simulated market resemble that of the real market in a much further degree: the return rate of the simulated market clusters not in the exact same way, but with a quite obvious trend and the "fat-tailed" distribution appears much clearer in the simulated market than in the original "Producer-Speculator" one.