一种数据处理的灰色预测新方法

A Method of Data Processing in Grey-forecasting

针对某些实验数据分散性强的特点 ,若直接用灰色GM (1,1)模型作预测 ,其误差较大。在此提出一种改进的灰色预测模型。首先将原始数据取λ次幂 (0 <λ <1) ,以降低它的分散性。由于λ可依具体情况取不同的值 ,所以比传统的对数据取对数更灵活 ,收敛性更好。然后再用GM (1,1)模型可以得到较好的预测结果。

For the characteristics of strong scatterence of some testing data,if a prediction is made from grey-forecasting model (GM(1,1) ),prediction value would get a rather big error compared with the real result,so a kind of new method to simulate the test was put forward.The data was firstly processed into powerlogarithmic series so as to reduce the scatterence of data, and then the prediction was carry out by the use of grey-forecasting model. For λ could choose different value according to idiographic instance,the accumulated series obtained from testing date possessed more better agility and convergence comparing with powerlogarithmic series. Therefore, a better result of prediction could be obtained by adopting the GM(1,1) model.