Abstract: The minimum spanning tree(MST) algorithm of complex network theory is combined with rolling windows analysis. In the context of the global financial crisis in 2008 and the shock of Chinese Stock Market in 2015, the daily closing prices of the Shanghai Stock Exchange(SSE) A-share market from 2003 to 2017 are chosen as the empirical data to construct the dynamic correlation network of China's stock market. From the perspectives of network node removal and edge-to-edge variation, the dynamic evolution and market robustness of Chinese Stock Market during different time periods are discussed. The relevant data is derived from the Wind. The results show that the whole network structure is loose before and after the financial crisis, while it becomes very compact during the financial crisis and stock market turmoil, and the degree distribution and influence-strength distribution of nodes are more concentrated. This means that during the financial crisis and the stock market turmoil, the influence of Systemically Important Financial Institutions will continue to improve and the changes in their stock prices will have a more significant impact on other companies. Therefore, from the perspective of policy formulation and supervision, relevant departments can focus on injecting liquidity into such companies during periods of financial instability, and then they will drive the trend of the entire market and make policy implementation more rapid and effective; different types of events have different effects on the connectivity and robustness of the stock market network. Regional events (stock market turmoil) have a greater impact on the market than global events (financial crisis); in the short-term, the financial crisis and the stock market turmoil have not caused much impact on the robustness of the entire network edge. Throughout the study period, the survival rate of the network has remained at a high level. But as the step size increases, the survival rate of the network drops sharply, especially if the step size is greater than three weeks (15 trading days), which means that the long-term robustness of the overall stock market network-related structure is declining. This paper can help to deeply investigate the structure characteristics and market status of Chinese Stock Market. At the same time, it has practical guiding significance for risk management of policy supervision and adjustment of portfolio strategy by market participants. In addition, the research can also be used to quantitatively measure the dynamic topology structure and robustness of other network-based financial systems.

Key words: financial crisis, stock market turmoil, complex network, stock market, dynamic evolution, robustness, systemically important financial institutions