GM(2,1)模型预测公式的改进研究  

Prediction Formula's Defect of GM(2,1) and Its Improvement

作　　者：赵新蕖[1] 陈红林[1] 

机构地区：[1]西北工业大学电子信息学院,西安710072

出　　处：《武汉理工大学学报》2006年第10期125-127,131,共4页Journal of Wuhan University of Technology

基　　金：航空支撑科技基金资助项目(02C53005)

摘　　要：从GM(2,1)模型的构造原理出发,认为其存在2个缺陷:Z(1)和X(1)预测公式混淆和强行定义边界条件。提出2种改进方案:用X(0)直接预测法代替Z(1)和X(1)预测后累减来预测X(0);用最小二乘的二次参数估计代替由边界条件确定白化微分方程的常数。计算示例表明,新方法具有更好的效果。From the structure principle of GM（2, 1 ）, two defects were pointed out ： confusing the prediction formulas of Z^（1） and X^（1） and being forccd to define boundary-conditions. Further more, two improvements were proposed： using directly predicting to X^（0） take place of the inverse-accumulating prediction, and using parameters estimated by least square to instead of that computed by boundary-conditions. Actual examples show that the new method reaches better result.

关 键 词：GM(2 1):X^(0) 直接预测 最小二乘估计 

分 类 号：N945.12[自然科学总论—系统科学]