中文摘要：

指数平滑法是应用广泛的时间序列预测方法之一，但在传统方法中其相关系数的确定具有主观性，因此，其预测结果往往偏差较大。本文对传统指数平滑法进行改进，将其参数动态化，使得模型随预测过程自动更新，从而保证了预测的实时性、客观性。以最小预测误差平方和（SSE）为优化目标建立动态指数平滑参数和初值的优化模型，并通过迭代优化法求解。通过动态指数平滑模型，传统方法的一些缺陷，如模型参数选取的主观性、易导致预测偏差等被有效解决。预测实例表明，新方法优于传统指数平滑方法。

英文摘要：

The exponential smoothing prediction methods have extensive application, but the choice of the exponential smoothing coefficients mainly depends on user＇s experiences, therefore, the predicted results are not reasonable. This paper presents algorithm to improve the performance of traditional exponential smoothing method, to define the dynamic parameters, and to adapt the changing of the prediction process on itself. The optimal model of the dynamic smooth parameter and initial value, aiming the square sum of errors （SSE）, are established, through which the corresponding dynamic parameters can be gotten impersonally. By dynamic exponential smoothing prediction method, some questions in traditional one, such as the model parameters are determined subjectively, lead to a deviation easily and so on, are resolved effec tively. The experimental results show that the presented method is higher in precision, better in adapta- tion and superior to the traditional one.