Dynamic Properties Analysis Of Chinese Stock Market From A Complex Network Perspective

In this paper, we investigate the dynamics of Chinese stock market from a complex network perspective, based on daily fluctuations of all stocks during 1906 working day period from 2005 to 2012. In the network being constructed, each node is a stock,and each edge indicates the Pearson correlation coefficient or normalized mutual information of two stocks over a window of T days. The network evolves chronologically as the window slides in forward time at a ?T-days interval.By examining the variation of the network parameters as time elapses, we find that network constructed by threshold method based on correlation matrix could reflect the trend of Chinese stock market. We show that 1) Different from the scale-free property observed in US stock markets, the degree distributions of Chinese stock market networks cannot follow the power-law at some periods; 2) When the Chinese stock market experiences a bear market, the average degree is exceedingly large and the ratio of edges existing at two sequential networks is high.By analyzing portfolios of the large-degree stocks, we show that they can fit in with the HS300 index well before 2009, and far exceed after. The reason might be the stocks with larger degree could benefit more under positive policy. Moreover, we analyze the composition of large-degree stocks and propose a method to select a small number of constituent stocks fitting the stock index. Furthermore, we discuss how a network approach can be used to build a well-diversified portfolio. We find that investments in stocks that occupy central, highly connected in the Chinese stock network outperform after the intervention of positive policy.