A consensus framework for multiple attribute group decision analysis in an evidential reasoning context
A consensus framework for multiple attribute group decision analysis in an evidential reasoning context
复制标题
DOI：
10.1016/j.inffus.2011.12.002
复制
发表时间：
2014-05
期刊：
影响因子：
--
通讯作者：
Chao Fu;M. Huhns;Shanlin Yang
中科院分区：
文献类型：
--
作者：
Chao Fu;M. Huhns;Shanlin Yang
在群体决策分析中，通常通过修改专家评价和调整专家权重两种策略中的一种来达成共识。由于缺乏对不真实变化的关注和对专家评估的忽视，本文建立了一个共识框架将两者结合起来。共识框架是在证据推理环境中实现的。它可以有效地处理由于缺乏或限制有关问题领域的知识、经验和可用数据而导致的对特定属性的评估缺失（这些属性称为缺失属性）的情况。基于对缺失属性达成最大共识的思想和小组讨论产生的建议有助于专家对缺失属性进行有效的评估。此外，共识框架包含一个反馈机制，为专家提供指导，以加速趋同达成共识。反馈机制包括属性层、可选层和全局层三个层次的识别规则和一条建议规则。前者表明建议具体专家更新其确定的损害协商一致意见的评估，后者则为专家更新其评估提出适当建议。如果在连续两轮推荐生成和评估更新后仍未达成共识，则使用仍然在三个层次构建的优化算法来调整专家的主观权重，以便于收敛到共识。用共识框架解决了一个工程项目管理软件选择问题，论证了共识框架的具体实施过程、有效性和适用性。
In group decision analysis, consensus has usually been reached by one of two strategies, modifying assessments of experts and adjusting weights of experts. Due to lack of attention paid to the unauthentic change and neglect of experts’ assessments, respectively, this paper develops a consensus framework to combine them. The consensus framework is implemented in an evidential reasoning context. It can deal effectively with the situation of missing assessments on specific attributes (the attributes are called missing attributes), which may be caused by lack or limitation of knowledge, experience, and available data about the problem domain. The recommendations generated based on the idea of reaching the maximal consensus on missing attributes and group discussion help experts to give effective assessments on missing attributes. Furthermore, the consensus framework contains a feedback mechanism to provide guidance for experts in order to accelerate convergence to consensus. Identification rules at three levels, including the attribute, alternative and global levels, and a suggestion rule are involved in the feedback mechanism. The former indicates that specific experts are recommended to renew their identified assessments damaging consensus, and the latter generates appropriate recommendations for the experts to renew their assessments. If consensus is still not reached after two consecutive rounds of recommendation generating and assessment renewing, then optimization algorithms still constructed at three levels are used to adjust subjective weights of experts so as to facilitate convergence to consensus. An engineering project management software selection problem is solved by the consensus framework to demonstrate its detailed implementation process, validity, and applicability.