中文摘要：

根据时间序列的结构与特征，对GM（1，N）灰微分方程进行了建模机理分析，并用数值积分算法提出了基于Simpson公式的建立GM（1，N）预测模型的新算法．用平均相对误差对一些时间序列进行了模型的实证分析，发现新算法的拟合精度比原有算法有明显的改进，从而验证了该算法对一些时间序列的有效性．所提出的新算法是建立GM（1，N）预测模型时值得尝试的一个方法，对GM（1，N）预测模型的合理应用具有一定的现实意义．

英文摘要：

This paper mainly deals with the analysis on mechanism modeling of GM（1,N） grey differential equations, based on the structure and characteristics of time series. By the numerical integration method a new algorithm of GM（],N） forecasting model based on Simpson formula is proposed. Using the method of average relative error, we take empirical anMysis on some time series models, and then find that the new algorithm has more accurate fitting precision than the original one. And the validity of the new algorithm is verified for some time series. The new algorithm is worth to be tried on establishing GM（1,N） forecasting model. And the reasonable application of GM（I,N） forecasting model has a practical significance.