Modeling and analysis of target influence in social networks

Abstract

Nowadays, research on social networks has attracted a large amount of attention from both academic and industrial societies. To understand the diffusion process and guide viral marketing, it is of importance to model and then estimate the influence of a seed user on a target user, which is defined as target influence in this paper. In famous diffusion models like independent cascade model and linear threshold model, tremendous computational costs are usually required in estimating influence probability through simulation. In this paper, we adopt duplicate forwarding model, and propose two measurements for the target influence, which can be analyzed theoretically. The former is the average number of duplicates the target user receives, and the latter is the probability of the target user receiving at least one duplicate. We further find the former will approach infinity if the spread intensity exceeds some threshold, but the latter can be adopted without this constraint. We also seek to use the latter to estimate the influence probability in the independent cascade model, and find it achieves much better accuracy than other heuristic metrics. All results are verified through simulations in real-world social networks, and we believe approach proposed here can provide insights to solve the problems like target influence maximization and influence maximization.