Abstract: For the problem of how to aggregate group information effectively of large-scale interactive group evaluation, this paper takes relational network as the breakthrough point and a methodology is proposed. Firstly, the evaluators are regarded as nodes and a relational network matrix is constructed by calculating the positive and negative correlation coefficient between nodes. Secondly, the node center degree and subgroup condensation degree are designed to measure the importance of node and subgroup. And then, the consistency and stability of group assessment information is measured to determine interaction termination conditions. Finally, we introduce the phase weight to aggregate information and order the results. An example validates the feasibility and effectiveness of this method.

Key words: interaction, relational network, node center degree, subgroup condensation degree, group evaluation

摘要： 针对大规模交互式群体评价中如何实现群体意见的有效集结问题,以关系网络为切入点,提出了一种基于关系网络结构的群体评价方法。首先,将评价者视为网络节点,并通过计算节点之间的正负相关系数来构造关系矩阵;其次,设计节点中心度和子群凝聚度,以度量节点和子群的重要性;然后,测量群体评价信息的一致性和稳定性,以此确定交互终止条件;最后,引入阶段权重以集结交互阶段的评价信息,并对最终结果排序。算例验证了该方法的适用性和有效性。

关键词: 交互, 关系网络, 节点中心度, 子群凝聚度, 群体评价