不完全信息下基于随机抽样的群组评价法*
2015-07-30闫英等
闫英等



摘 要:针对不完全信息下的群组评价问题,借鉴D-S证据理论的表达方式和思想,通过计算专家评价意见间的相似度矩阵,得到专家意见的客观权重,以此为依据对不完全信息进行融合,根据融合后的信息构造评价对象的累积分布函数,采用Monte Carlo随机抽样法得到最终的评价值,对评价对象进行排序。实例分析表明新方法正确、有效,且能较好的保留专家原始评价信息,得到更为客观的评价结果。
关键词:不完全信息; 群组评价; 随机抽样; D-S证据理论; 证据融合
中图分类号:N945.25 文献标识码:A 文章编号:
A Group Evaluation Method Based on Random Sampling under Incomplete Information
YAN Ying1, SUO Bin2, YUAN Ming1, GAN Mi3
(1. School of Economics and Management, Southwest University of Science and Technology, Mianyang 621010;
2. China Academy of Engineering Physics, Mianyang 621900, China;
3. School of Traffic Transportation & Logistics, Southwest Jiaotong University, Chengdu 610031, China)
Abstract: Aiming at group evaluation problem under incomplete information, a new method based on random sampling was proposed. Drawing on the expressment and thoughts of D-S evidence theory, similarity matrix between each two experts opinions was calculated to determine the objective weights of experts opinions, with which imcomplete information were fused. Cumulative distribution functions of evaluation objects were constructed based on fused information. Monte Carlo sampling was used to get the final values of objects, and then objects were sorted according to these values. Example shows the effectiveness and correctiveness of the new method. With this method, raw information in experts opinions are retained, and the evaluation results are more obejective.
Key words: incomplete information; group evaluation; random sampling; D-S evidence theory; evidence fusion
引言
在实际的群组评价过程中,由于决策者自身知识和经验的局限性、时间的紧迫性,在面临复杂评价问题时,往往存在评价信息不完全的现象。例如,对于几个不同类型的项目进行投资风险评价,从中选择一项风险小、前景好的项目进行投资,不同的决策者可能仅对其中一两个项目涉及的行业比较熟悉,其他项目并不熟悉,但仍要做出评价,此时决策者往往倾向于对不熟悉的项目给出不完全评价信息。由于不完全信息评价问题有着重要的实际应用价值,近年来成为决策和评价领域的热点问题[1~11],其中基于D-S证据理论的方法受到了国内外学者的广泛关注[5~11]。文献[5]、[6]将D-S证据理论与AHP法相结合,通过AHP法求解判断矩阵的特征向量,将决策者的不完全信息转化为证据理论下的基本概率分配,进而采用Dempster规则对信息进行融合;文献[7]、[8]主要针对权系数不完全确定且方案的部分准则值位于两个语言等级之间的情况,提出了D-S证据理论融合决策信息、构造非线性规划模型的方法计算各方案的排序;……
