A novel consensus model for multi-attribute large-scale group decision making based on comprehensive behavior classification and adaptive weight updating
A novel consensus model for multi-attribute large-scale group decision making based on comprehensive behavior classification and adaptive weight updating
复制标题
基于综合行为分类和自适应权重更新的多属性大规模群体决策共识模型
DOI：
10.1016/j.knosys.2018.06.002
复制
发表时间：
2018-10-15
影响因子：
8.8
通讯作者：
Ding, Ru-Xi
中科院分区：
文献类型：
--
作者：
Shi, Zijian;Wang Xueqing;Ding, Ru-Xi
近年来，随着对双方同意的决策结果的需求大大增加，共识达成过程（CRP）受到越来越多的关注。在当前将电子民主和公众参与引入公共问题决策的趋势下，不同背景的决策者在试图达成共识时更容易遇到冲突，尤其是在多属性的大规模群体决策框架下。为了提高CRP的效率，人们提出了不同的共识模型。决策者提出的特定行为模式，例如不合作行为和少数意见，也在这些模型中受到严格监督。然而，并不是每种类型的行为都被明确定义并给予定向处理，这包括高权重集群的行为，这可能会严重偏差群体共识。在本文中，我们提出了一种新颖的 CRP 模型，称为基于 uninorm 的综合行为分类（UBCBC）模型，该模型具有增强的效率和合理性。首先，提出一种基于合作指标和非合作指标计算的行为分类模型，对三种修改行为进行分类。其次，使用统一聚合算子更新 CRP 下一次迭代中的决策权重，以奖励或惩罚集群的行为。此外，在统一聚合算子中引入了浮动中性元素，以对高权重集群进行更严格的监督。最后通过算例和数值模拟证明了该模型的高效性和可行性。
Consensus reaching process (CRP) has received increasing attention in recent years, as the demand for decision results with mutual agreement has greatly grown. With the current tendency to introduce e-democracy and public participation into decision making for public issues, decision makers from various backgrounds are more likely to encounter conflict when attempting to reach a consensus, especially under a multi-attribute large-scale group decision making framework. In order to improve the efficiency of the CRPs, different consensus models have been proposed. Specific patterns of behaviors presented by decision makers, such as non-cooperative behaviors and minority opinions, are also strictly supervised in these models. However, not every type of behaviors is specifically defined and given directed treatment, this includes the behavior of highly-weighted clusters, which may seriously bias group consensus. In this paper, we present a novel CRP model named uninorm-based comprehensive behavior classification (UBCBC) model with enhanced efficiency and rationality. First, a behavior classification model based on the calculation of a cooperative index and a non-cooperative index is proposed to classify three kinds of modification behaviors. Second, decision weights in the next iteration of the CRP are updated using a uniform aggregation operator to reward or penalize the behaviors of clusters. Furthermore, a floating neutral element is introduced into the uninorm aggregation operator to lay stricter supervision upon highly-weighted clusters. Finally, an illustrative example and a numerical simulation are implemented to prove that this model is of high efficiency and feasibility.