Detection of Microblog Overlapping Community Based on Multidimensional Information and Edge Distance Matrix

摘要

The traditional community detection algorithms are often based on the network structure, without considering the unique characteristics of weibo (microblog) network. In this paper we proposed a weibo overlapping community detection algorithm, called MIEDM. It takes weibo network as research object and is based on multidimensional information and edge distance matrix. First, we established the weighted network topology graph integrating the weibo multidimensional information such as weibo user relationship, user behavior, weibo theme, and geographic location. Second, based on the edge-node-edge random walk model, we constructed the edge distance matrix. The matrix not only considers the distance of adjacent edges but also the distance of non-adjacent edges. Then, we improved the existing density peak clustering algorithm, and employed the improved algorithm to identify the initial communities with the edge distance matrix considered. In addition, the initial discovered communities arc merged and optimized according to the modularity increment. Final, the results of experiments on the weibo network and real networks show that this algorithm yields higher accuracy, stability and generality.