摘要
针对广义最大熵回归方法的建模效果问题,尤其是模型中未知参数和误差项支持空间选择的不确定性问题,该文剖析了该方法的建模过程,并通过两个实例将该方法与其它建模方法的回归效果进行了对比分析。结果表明:广义最大熵回归方法的预测精度与解释能力优于最小二乘法和偏最小二乘法以及主成分方法;在先验信息缺乏的情况下,参数支持空间越大越好;误差项支持空间应在3σ与4σ之间。
Aiming at generalized maximum entropy （GME） regression effect and especially the indetermination of the choice of support space of parameter and error in the model, the modelling process of GME regression method is analyzed, and its regression effect is compared with others＇ effects through two cases in this paper. Results show that the forecasting precision and the explaining ability of GME method is higher than the least square method, the partial least square method and the principal component analysis. The larger the support space of parameter is, the better, under the lack of prior information, and the support space of error should be between 3σ and 4σ.
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2006年第6期793-796,共4页
Journal of Nanjing University of Science and Technology
关键词
回归方法
广义最大熵
最小二乘法
偏最小二乘法
regression method
generalized maximum entropy
least square method
partial least square method
