Research on community evolution based on node influence and multi-attribute fusion

Abstract

For the problem of low discrimination accuracy of evolutionary events in dynamic social networks, a community evolution model (EMNI) based on node influence and multi-attribute fusion is proposed. Firstly, the topological structure information of nodes is obtained by random walk and local clustering coefficient, and the influence of nodes is evaluated according to the topological structure of nodes. Secondly, in order to improve the accuracy of discriminating community similarity, a community similarity discrimination method based on multi-attribute fusion is proposed. The model EMNI combined the characteristics of community stability and community difference, and redefined seven evolutionary events. Finally, the effectiveness of the EMNI model in identifying community evolution events is verified on different data sets. The experimental results show that the EMNI model is better than GED, PECT and SGCI, which is able to identify more evolutionary events and the distribution of events is also more balanced.