Dynamic Community Detection Based on Similarity of Social Network Nodes

Abstract

As an important tool for analyzing the structure and composition of social networks， predicting nodes connections and the evolution process, community detection is widely used in recommendation systems, public opinion detection and other fields. Most traditional dynamic community detection algorithms rely too much on the previous community structure, ignoring the dynamic characteristics of node structure and attributes changing with time in the network, resulting in poor dynamic community detection results. Therefore, this paper proposes a dynamic community detection algorithm based on the similarity of social network nodes (DCDSN). The algorithm calculates the static similarity between nodes through structural and attribute similarity according to the three influence factors of direct influence, common neighbor and exclusive neighbor. The algorithm uses the influence of historical network information on current network information to obtain the dynamic similarity between nodes, establishes a weighted dynamic social network, then uses Louvain algorithm to achieve dynamic community detection. Experiments show that DCDSN algorithm has higher community detection quality than other community detection algorithms on Enron dataset.