基于多准则的组合预测模型权重研究及其应用

来源期刊:中南大学学报(自然科学版)2015年第5期

论文作者:黄仁东 张海彬 杨志辉 周扬 张攀

文章页码:1778 - 1786

关键词:组合预测;证据理论;权重调整系数;可信度;证据精度;自冲突系数

Key words:combination forecast; evidence theory; weight adjustment coefficient; credibility; evidence precision; evidence contradiction

摘    要:针对传统组合预测模型大多是通过建立单一准则方程进行优化,而没有更好地考虑各单一模型之间互支持信息带来的不确定性问题,建立基于多准则的组合预测模型权重确定算法。首先,通过建立区间数模型构建样本区间距离并进行相关折算归一化生成样本的基本概率分布BPA(basic probability assignment),作为单一预测模型的初级权重;然后,通过对D-S证据理论进行改进,建立证据可信度、证据精度和证据自冲突系数3个准则分别用于评价单一模型自身精度及其之间互相支持信息,通过对3个准则排序得到综合排序值作为单一模型初级权重的权重调整系数;最后,综合多时刻数据归一化后确定单一模型的最终权重用于组合预测。研究结果表明:经过权重调整后的组合预测精度得到显著提高,且经过调整系数R调整后的不变权组合预测模型最优。

Abstract: Considering that most of traditional combination forecast models are established by criterion of single equation regardless of information between individual model which implicits lots of uncertainty, a weight determination algorithm of combination forecast model was presented based on multi-criteria information. Firstly, interval numbers model was built to get the sample BPA matrix as the primary weight of the single models by a series of distance calculations between the true data and the predictive value. Then, three criteria, i.e. evidence credibility, evidence precision and evidence contradiction, were set up to evaluate the precision of individual model and information between them. By sequencing the above three criteria, a composite sort value called weight adjustment coefficient was generated to adjust the primary weight of the single models. Finally, the final weight was determined by normalizing the multi-time weight data used for combination forecast. The results show that the precision of the method is high, one fixed weight combination forecast model adjusted by weight adjustment coefficient R is the best.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号