Opinion Leaders Discovering in Social Networks Based on Complex Network and DBSCAN Cluster

Abstract

The opinion leaders play an important role in the process of network public opinion spreading. In order to quickly and efficiently discover the opinion leaders, this paper analyzes the characteristics of complex networks in social networks and proposes density-based spatial clustering of applications with noise algorithm based on local community detection method. With Sina micro-blog user as the research object, the feature vectors of opinion leaders are extracted as the training set, then the characteristic means of the subclass are obtained, from which the user groups with the community opinion leader characteristics can been identified. Finally, DBSCAN algorithm is compared with the K-means algorithm and the average path length difference algorithm by using the same data set. The experiment results show that DBSCAN algorithm can be more accurate and more effective to find community opinion leaders.