This paper uses the methods of complex science,and studies the statistical properties of probability distribution and memory effect of the recurrence intervals,defined as the intervals between extreme events in financial markets.Relatively accurate analysis is performed by using various goodnessoffit tests and memory effect detecting methods.The application of the recurrence interval analysis in financial complex systems is systematically introduced,including the empirical studies of recurrence intervals of volatilities,realized volatilities,returns and trading volumes.Based upon this study,we perform a primarily risk estimation for the stock market.Furthermore,an order driven model is introduced.The dynamics of the recurrence intervals between extreme events is studied by mimicking the order submission process of the investors in stock markets.This work may provide a theoretical foundation for the risk estimation and avoiding of extreme events in the Chinese stock market.