中文摘要：

本文基于极值理论，以沪铜、大连大豆和郑州硬麦期货为研究对象．应用最大信息熵原理、广义极值分布和K-S拟合检验，实证分析了我国商品期货收益率极端波动的时间间隔的统计分布特征。引入离散的时间间隔来刻画极端波动特征在学术上是个新的尝试。实证结果表明：极端波动的确定是与阈值紧密联系的：我国商品期货收益率的极端波动间隔时间服从广义极值分布；期货收益率极端波动时间间隔序列具有集聚性；运用条件期望值——波动模型预测极端波动间隔是可行的。这些成果为监管层度量极端波动强度和控制风险提供了有效的方法。

英文摘要：

This paper selects futures contracts of copper in SHFE, hard wheat in ZCE and soybean in DCE as samples and, by means of principle of maximum entropy, generalized extreme value distribution （GEV） and K-S goodness-of-fit test, makes an empirical study on the characteristics of statistic distribution of time interval between extreme volatilities of return rate series of commodity futures in China. It is a new academic attempt to introduce the discrete time interval to portray the characteristics of extreme volatility. The empirical results show that the determination of extreme volatility is closely related with the threshold; time interval between extreme volatilities of futures return rate in China submits to generalized extreme value distribution; there is character of clustering in time interval series between extreme volatilities of futures return rate; it is feasible to predict the time intervals of extreme volatilities by means of condition-expectation-value-volatility model. These results provide efficient methods for supervisors to measure the intensity of extreme volatility and to control risk.