Key Nodes Evaluation Method Based on Combination Weighting VIKOR in Social Networks

Abstract

Evaluation of key nodes is a hot issue in social networks. Existing research primarily evaluates the importance of nodes in social networks based on centrality metrics, neglecting the node’s own attributes. After analyzing the topology attributes and the basic attributes of nodes, this article proposes a key nodes evaluation method for social networks, which is based on analytic hierarchy process (AHP) and improved Vise Kriterijumska Optimizacija I Kompromisno Resenje (VIKOR), termed AE-VIKOR. Considering global attributes, local attributes, and positional attributes of nodes, three evaluation metrics are constructed. The subjective and objective weights are computed by AHP and entropy weight method, respectively. The comprehensive weights of metrics are determined by combination weighting method based on square sums of distance. Due to the excessive weight of specific metrics and excessive difference in data distribution, the computation of individual regret value depends too much on a single metric in VIKOR method, individual regret value is optimized by weighted sum of closeness between the scheme to be evaluated and the negative ideal scheme. Multimetric evaluation schemes are ranked to achieve the evaluation of key nodes. Experiments on two real social network datasets show that the key nodes evaluated by AE-VIKOR have stronger information spread ability and more fans than the ones of the existing methods. In addition, the validity of the three metrics and the two improvements on the VIKOR method are verified by ablation experiments.