【摘要】 在传统的灰色预测模型中 ,白化微分方程的初始条件和辨识参数一旦确定便不再改变 ,这不太适合于动态及长期的数据预测 .针对这种情况 ,提出了一种能动态选择最优初始条件及相应辨识参数的新型灰色 GM( 1 ,1 )预测模型 .该模型先对每个预测初始值都用 GM( 1 ,1 )进行预测 ,并计算平均相对预测误差 ,使平均相对预测误差最小的那个观测值即为微分方程的最优初始条件 .然后将用此初始条件预测的新信息数据替换原来最老的那个数据 ,算出新的参数 ,即动态求解辨识参数 .将此模型用于时程数据的预测中 ,取得了较好的预测效果更多还原

【Abstract】 In original GM(1,1) model, initialization condition and identifying parameter no longer change as long as they are confirmed. However, it is not fit for dynamic and long data prediction. As a result, a new prediction model, which can choose the best initialization conditions and dynamically confirms identifying parameters, is proposed in the paper. The model calculates sums of errors according to every measure value, and considers the value, which makes the sums of errors minimum, as the best initialization condition. And then, new identifying parameters are calculated by replacing the oldest data with new data. The model is applied to prediction of time-displacement data and the results are satisfactory.更多还原