A cascade information diffusion based label propagation algorithm for community detection in dynamic social networks

Abstract

One of the most important topics in social network analysis is community detection in dynamic social networks. A variety of approaches exists for detecting communities in dynamic social networks, among which the label propagation algorithm (LPA) is the well-known approach. This approach has made remarkable performance, but still has several problems. One of the difficulties of this approach is the new nodes added to the social network graph in the current snapshot has a very slight chance of creating new communities. In fact, these nodes fall under the influence of existing communities. This drawback decreases the accuracy of community detection in dynamic social networks. We propose a new method based on label propagation approach and the cascade information diffusion model in order to solve this difficulty. Here, the newly proposed method, Speaker Listener Propagation Algorithm Dynamic (SLPAD), Dominant Label Propagation Algorithm Evolutionary (DLPAE) and Intrinsic Longitudinal Community Detection (ILCD) on real and synthetic networks are implemented. The findings indicate that the modularity and Normalized Mutual Information (NMI) and also F1AVG of this proposed method is considerably higher than the earlier available methods in most datasets. Therefore, it can be concluded that the proposed method improves the accuracy of community detection in comparison with other available methods.