考虑网络交互影响效应的评价者权重分配方法.pdf

考虑网络交互影响效应的评价者权重分配方法 李永立 吴 冲 张晓飞 哈尔滨工业大学 管理学院，哈尔滨 150001 摘要：确定评价者的权重是评价理论与方法重要的一步。提出了基于网络博弈框架的新的评价者权 重分配方法，并就解和参数的性质进行了详尽地讨论与验证。将评价者视为网络节点，评价者的评 价行为形成评价者之间的联系，构成网络的边，根据评价者的评分信息定义他们之间的“合作”和 “冲突”矩阵构成网络的权重矩阵，建立了网络博弈的模型，其最优解即为权重值。给出了模型最 优解的求解方法，并证明了该最优解是 Nash 均衡的内点解，并且是这一问题唯一的 Nash 均衡解。 分析了最优解与网络中心性测度之间的联系，较为本质地分析了最优解在网络科学中的含义，将决 策理论与网络科学的知识关联起来。通过数学证明和仿真分析，揭示了模型中参数的含义，其对于 解的“保序性”和“稳定性”两个优良性质是此消彼长的关系，提出了根据数据集特征选择参数值 的相机抉择的管理思想；并在真实的数据集上进一步做模型的应用分析。本文提出的考虑网络交互 影响效应的权重分配方法具有实用性，能够平衡解“保序性”和“稳定性”两个良好属性，具有理 论意义和实践价值。 关键词：评价理论与方法; 优化模型；网络分析；交互效应；网络博弈；权重分配；仿真 中图分类号：F724.6 文献标识码：A An Evaluator Weight Allocation Method Considering Network Peer Effect LI Yongli WU Chong ZHANG Xiaofei School of Management, Harbin Institute of Technology, Harbin 150001 Abstract ：Determining the weights of evaluators is an important step in evaluation theory and method. This paper proposed a new evaluator weight allocation method based on network game and besides detailed discussions and validations of the solution and parameters were given. This paper considered evaluators as network nodes, and evaluators’ evaluations as the links between themselves, which made up the edges of the network. Based on evaluators’ rating information, this paper defined the “Cooperation” and “Confliction” matrix between them as the weight matrix of the network. The establishment of a network game model is finished and the optimal solution is solved as the weight values. We also proved the optimal solution is Nash interior equilibrium solution and the only Nash equilibrium solution of this problem. Furthermore, this paper analyzed the relationship between optimal solution and network-centricity measure, especially analyzed the meaning of optimal solution in n