The Dynamic Influence Graph Model on Mobile Datasets

Abstract

With the rapid development of mobile technologies, more and more people are equipped with smartphones. It is possible for scientists to collect and analyze mobile data efficiently. Mobile data contain rich semantic as well as topological information. Rich information can be inferred from these data such as social influence among different nodes in mobile social network. However, it is difficult to estimate the strength of social influence due to the characteristics of inherent dynamic and large scale of mobile social network. In this paper, a Dynamic Influence Graph (DIG) model is proposed which utilizes temporal information in a topological perspective, and an efficient algorithm is proposed based on the DIG model. The proposed algorithm can calculate social influence between any two nodes in a given mobile social network stream segment, and takes edge weights, node connectivity and temporal information into consideration. Experimental results with a real mobile social network dataset show that the proposed approach can infer social influence and achieve a-state-of-the-art accuracy (82-86%) efficiently and automatically.