大规模群体决策中最小成本最大满意度模型研究

Optimization of the Minimum-Cost Maximum-Satisfaction Model in Large-Scale Group Decision-Making

摘要: 近年来，基于社交网络分析的大规模群体决策问题受到了广泛关注，但随着数字经济的发展，大群体决策的场景不断拓展，决策问题本身也呈现出高度的复杂性与不确定性。在此背景下，决策群体的构成发生了显著变化，专家的异质性日益凸显，这会导致专家在面对不同的决策时会有不同的共识目标。但在现有研究中很少有文章从专家的异质性出发研究多层次的优化目标。因此本文重点研究基于局部搜索算法的双层目标共识优化方法，首先我们利用一种新的社区检测方法——局部搜索算法挖掘专家的局部信息识别出社区结构，并构建多层网络模型来清晰表征专家间的动态交互关系。其次构造了最小成本最大满意度的双层优化模型填补了该领域的空白。文中还对模型进行了案例应用和敏感度分析，结果表明：本文提出的方法能够取得较优的结果，并且所提出的模型具有一定的理论指导意义。最后本文对于后续工作进行了展望。

Abstract: In recent years, large-scale group decision-making problems based on social network analysis have received widespread attention. However, with the development of the digital economy, the scenarios of large-scale group decision-making continue to expand, and the decision-making problems themselves exhibit a high degree of complexity and uncertainty. In this context, the composition of decision-making groups has undergone significant changes, and the heterogeneity of decision-makers has become increasingly prominent. This leads decision-makers to have different consensus goals when facing different decision-making scenarios. Nevertheless, existing research rarely explores multi-level optimization objectives from the perspective of decision-maker heterogeneity. Therefore, this paper focuses on a bi-level objective consensus optimization method based on a local search algorithm. First, we use a novel community detection method—the local search algorithm—to mine the local information of decision-makers and identify community structures. A multi-layer network model is then constructed to clearly characterize the dynamic interactions among decision-makers. Second, to address the gap in this area, a bi-level objective optimization model is developed to achieve minimum cost and maximum satisfaction. A case application and sensitivity analysis of the model are also conducted. The results demonstrate that the proposed method yields superior outcomes and provides certain theoretical guidance. Finally, future research directions are discussed.