论文部分内容阅读
首先将证券市场运动用局部多项式趋势模型进行平滑 ,然后分别用 AR模型和GARCH模型考虑序列之间自相关性和波动的变化性。参数的条件最大似然估计应用了状态空间模型的卡尔曼滤子递推和 GARCH模型的条件方差递推 ,模型阶数的选取应用了Akaike的最小化信息矩阵方法。计算实例表明了这种组合方法预测能力的优越性。
Firstly, the local market polynomial trend model of security market is smoothed. Then, AR model and GARCH model are respectively used to consider the variability of the autocorrelation and volatility between the sequences. The conditional maximum likelihood estimation of parameters applies Kalman filter recursion of state space model and conditional variance recursion of GARCH model. Akaike’s minimization information matrix method is applied to the selection of model order. The calculation example shows the superiority of this combination method in forecasting ability.