中文摘要：

在收益率波动层次结构理论上构建了一种新的波动级串结构模型，通过对基元分割概率中两个关键控制参数不同设定的Monte-Carlo仿真分析发现，分割数取随机整数能够显著地改善模型的拟合优度，自相关函数也表现出类似经验数据衰减的动力学特征，随机控制变量选取normal、log-normal，possion和t分布不能进一步改善基于均匀分布级串模型的拟合效果。此外，选取的最优波动级串模型还能够很好地复制出2〈α〈4的负幂律衰减和多重分形特征，表明该级串模型能够作为经验数据的可靠拟合模型。

英文摘要：

A new volatility cascade model is constructed based on the hierarchy of volatility. The Monte-Carlo simulations of changing the combinations of two key control variables observe that for the partition taking the random integer, the model can increase the fitting effect significantly, and that the autocorrelation function can also exhibit the similar decaying dynamics as the empirical. However obeying the other four different distributions, such as normal, log-normal, passion and student-t, of random control variable can＇t further improve the original fitting. In addltion, the optimal cascade model reproduces well another two typical features often shown in empirical data, negative power-law decaying with exponent 2〈α〈4 and multifractality. These results show that this cascade model can be used as the reliable model to explain the empirical phenomena.