Estimating user influence ranking in independent cascade model

Abstract

Nowadays, hundreds of millions of people use social networks to express their opinions and communicate with their friends. It is of importance to model and estimate the user influence in social networks. Since most studies perform Monte Carlo simulation to evaluate the user influence in the independent cascade model, which leads to tremendous computational costs, we introduce a duplicate forwarding model to characterize the diffusion process in social networks, and analyze the user influences below and above the diffusion threshold theoretically. After getting the user influence ranking, we propose a Spearman-like correlation coefficient to measure the correlation between two rankings, and find the analysis results from the duplicate forwarding model achieve much better accuracy than the measurements degree, betweenness, k-core and PageRank in estimating the user influence ranking in the independent cascade model. This approach can provide insights in modeling and estimating the influences of social network users, and can be easily extended to estimate the influence ranking for different seed sets in the problem of influence maximization.