中文摘要：

GARCH类模型和扩散类模型是常用于研究利率的两类模型。运用这两类模型对我国银行间国债隔夜回购利率进行实证研究，结果表明短期利率波动存在明显的非对称性。在预测均值时两类模型表现较差，而在预测波动率水平时，GARCH类模型具有较好的预测效果，其中数GARCH—M和TGARCH表现最好，虽然扩散类模型在预测方面表现不好，但由于具有马尔科夫性，所以在衍生产品定价上具有明显优势。

英文摘要：

GARCH - typed patterns and decentralizing - typed patterns are two commonly used patterns for research on interest rate. A study is made on interbank T- bond overnight repo interest rate in China by using these two patterns. The result suggests that there is apparent asymmetry for short -term interest rate fluctuation. When antici- pating the averages, performance of these two patterns is rather worse, and when anticipating the levels of fluctua- tion, GARCH- typed patterns show a better anticipating effect, with GARCH -M and TGARCH illustrating the best effect. Though decentralizing - typed patterns illustrate worse anticipating effect, they are with the nature of Markov, hence having an apparent advantage in derivative pricing.