Prediction of link evolution using community detection in social network

Network evolution is one of the emerging research directions in the field of social network analysis, where link prediction plays a crucial role in modeling network dynamics in social networks. Link prediction has attracted a lot of attention of network engineers in developing several applications. In this paper, an effort has been made to model the network evolution through link prediction termed as LP-CD (Link prediction through community detection). We have leveraged the set of existing communities in the network for link prediction. Different community detection algorithms have been implemented to identify the dense subgroups in the network, which are further used in predicting future links. After identifying the dense subgroups in the network, the non-existing links inside the subgroups are identified using the global path-based similarity measures. The intuition of the proposed model is that two users are more likely to form a relationship in the future, if they belong to the same community and less likely to establish a relationship, if they belong to different communities. An extensive comparison of various existing models with LP-CD has been made by considering four real-world and three synthetic network datasets. The results show that the LP-CD outperforms other approaches for link prediction.

Automated summary

This paper introduces a link prediction model called LP-CD to simulate network evolution. It leverages existing communities in the network for link prediction and uses global similarity measures to identify non-existing links in the future. Experimental results show that LP-CD outperforms other approaches for link prediction.