论文部分内容阅读
时间序列参数估值的时变性与预报模型的定常性之间的差异,使得预报精度随着预报步数的增加而显著降低。基于此,构建基于卡尔曼滤波和自适应卡尔曼滤波的AR模型以实时更新其参数估值,取得了较好的拟合效果并提高了预报精度。
The difference between the time-varying estimation of time-series parameters and the regularity of forecasting model makes the forecasting accuracy decrease significantly with the increase of forecasting steps. Based on this, an AR model based on Kalman filter and adaptive Kalman filter is constructed to update its parameter estimation in real time. The fitting result is good and the prediction accuracy is improved.