Dynamic Relationship Network Analysis Based on Louvain Algorithm for Large-Scale Group Decision Making

Abstract

In most existing large-scale group decision making (LSGDM) problems, the relationships between decision makers (DMs) are usually ignored or regarded as static. However, in many cases, the results of LSGDM are dynamically influenced by the relationship between group members. To address this issue, a dynamic relationship network analysis method based on Louvain algorithm is proposed in this paper. First, each DM could be considered as a node to construct a relationship network, which dynamically change the individual opinion by the definition of correction index to eliminate subjective factors. Second, the node centrality and subgroup cohesion are defined and the Louvain algorithm is used to divide DMs into several subgroups to measure the importance of each node and subgroup. Then, the termination conditions of the discussion are determined by measuring the consensus and stability of the group decision information. Moreover, stage weight function is defined to assign weights to discussions at different stages and obtain the final results. An illustrative example is provided to prove the feasibility of the proposed model. Sensitivity analysis is given to show the stability of correction index and stage weight function. Finally, the comparative analysis is performed to illustrate its feasibility and effectiveness of the method.

Keywords

• large-scale group decision making
• consensus reaching process
• dynamic relationship network
• Louvain algorithm
• node centrality
• subgroup cohesion