Abstract: In recent years, extreme events such as continuous soaring and plummeting have occurred frequently in Chinese stock market, and the management of extreme risks in the stock market has been severely challenged. In order to explore the relationship between extreme events and the impact of Investor Sentiment on the extreme returns of the stock market. In this paper, the Process of the Marked Self-exciting Point Process (MSEPP) is used to describe the clustering and short-term dependence of the extreme return series of stock index, and the homogeneous Poisson process with fixed intensity λ described by the traditional Peaks Over Threshold(POT) model is extended to the non-homogeneous Poisson process with time-varying intensity function λ(·). Using the method of risk preference index, the Equity Market Sentiment Index(EMSI) of Chinese stock market is synthesized based on the CSI300 Index Components. Taking EMSI as one of the explanatory variables of the strength function λ(·), the MSE PP-EMSI model is further constructed to predict the extreme risk Outbreak Probability of CSI 300 index, Shanghai composite index and Shenzhen component index during the stock market crash in 2015, and to measure the dynamic Expectd Shortfall(ES) risk of these indexes from June 9, 2017 to March 30, 2018. The empirical results show that the stock indexes of Shanghai and Shenzhen stock markets have plummeted continuously in the short term, and investors' extreme negative emotions will aggravate the violent turbulence of the stock market. When considering the impact of investors' emotions on extreme risks, MSEPP-EMSI model can effectively improve the probability prediction accuracy and ES prediction accuracy of extreme risks. The conclusion of this study reveals the Self-exciting effect of extreme risks in Chinese stock market, and explores the impact of investor sentiment on stock market returns, expanding the research in related behavioral finance fields. At the same time, it can effectively guide the trading behavior of stock market participants, enhance the risk management level of institutional investors in the face of extreme risks, and provide a basis for government regulatory authorities to formulate policies.

Key words: extreme value theory, marked self-exciting point process, stock market sentiment index, risk probability forecast, ES measure