摘要
为解决多层递阶方法的预报效果不稳定 ,特别是预报因子之间存在较大量级差异时预报不稳定的问题 ,给出了一种改进的多层递阶预报方法———多层递阶回归分析方法 .该方法将原模型中的各项看作回归变量作线性回归 ,再以回归系数与预报因子的乘积作为对原预报因子的修正变量 ,然后进行多层递阶预报 .该方法集多层递阶和回归分析两者的优点 ,既能较好地体现高相关因子在预报模型中的重要作用 。
An improved way of analysis, namely multi-degree regression analysis, is proposed to better the uncertain effects of multi-degree prediction analysis after time sequence, especially those while there is a larger diversity in terms of quantity. In this new approach, the variants in the original mode are taken as regression variants which functions as linear regression, and then the product of the regression coefficient and the predictive factor is taken as the revised variant of the original predictive factor. Thus a multi-degree prediction is made. This method has the merits of both multi-degree and regression analysis, reflecting the important function of the related factor in the prediction mode and greatly adapting to the system of sequential variance.
出处
《哈尔滨工业大学学报》
EI
CAS
CSCD
北大核心
2002年第3期436-439,共4页
Journal of Harbin Institute of Technology
