摘要
估计指数模型参数通常采用最小二乘法,而在检验模型精度时又经常采用平均相对误差绝对值,因而所得模型必然不是最优模型.针对这一问题,在平均相对误差绝对值最小的准则下,给出了一种优化GM(1,1)模型参数的方法,该方法具有全局收敛性,只要选取较好的初值,总能得到更好的估计值.在工程、管理中的两个应用实例表明了此方法的正确性和可行性,且模型精度高于传统方法.
Exponential model parameters are estimated usually by least square method,but the accuracy of the model is often tested by the mean absolute percentage.Thus the modeling is unsatisfactory according to the criterion of the minimum mean absolute percentage errors.On the basis of the criterion,the algorithm to optimize GM（1,1） model parameters under the criterion of minimizing mean absolute percentage error is proposed.This algorithm is global convergent.When selecting a good initial approximation,a better model parametersr estimated value is obtained. Practical applications to engineering and management show that this method can enhance the accuracy of model.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2009年第10期29-31,共3页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(60674071)
温州大学研究生创新基金资助项目(YCX0810)
温州大学教改研究项目(08JG06A)
