Research on Communities Detection in Social Network

Abstract

During the evolution of social network, there is a social network phenomenon that small communities also become important. Generally, each community has its own characteristics of internal correlation and relation. Accurate division of whole social networks into multiple small communities may help improve the quality of social network services as whole. With the comparison among substantial community detection algorithms, we present a Label Propagation Algorithm (LPA), which has proven to be more efficient for large scale community detection and widely used. Random (node) access orders within the algorithm severely hamper its robustness, consequently, and the stability of the identified community structure. In this paper, we propose a Precedential Label Propagation Algorithm (PLPA) which counteracts for the introduced randomness by increasing propagation preference. The experiment results verify the PLP is more robust than LP.