中文摘要：

以沪铝期货市场为研究对象，针对金融市场的有偏性、尖峰厚尾性，结合条件极值理论与SKST分布刻画金融市场的极端风险。同时运用滚动时间窗口方法对不同波动率模型进行样本外动态VaR预测。鉴于传统的回测检验无法有效判断不同波动率模型风险测度效果的优劣性，本文引进一种新的风险检验方法——MRC—SPA检验，实证结果显示EVT有效提高了GARCH模型的样本外动态VaR预测精度，其中GARCH—SKST—EVT—POT模型以较小的市场风险资本实现风险规避，预测效果最优。

英文摘要：

Based on the characteristics of aluminum futures market in China, we introduce the Skewed-t distribution combined with conditional extreme value theory to describe the thick tail, skewness and volatility clustering and use rolling time window for different volatility models to forecast the dynamic out-of-sample VaR. Then we choose the MRC-SPA test, a new test for risk, to cover the inefficiency caused by traditional tests on different volatility models. The main empirical results show that： EVT effectively improve the prediction accuracy for dynamic out-of-sample VaR and can achieve a smaller deal of regulatory capital loss, what is more, GARCH-EVT- Skewed-t is the best prediction model.