Abstract
I. Introduction
II. Related Work
III. OI-Based Method
IV. Experiments and Results
V. Conclusion
Authors
Figures
References
Abstract
Online social networks (OSNs) are interaction platforms that can promote knowledge spreading, rumor propagation, and virus diffusion. Identifying influential users in OSNs is of great significance for accelerating the information propagation especially when information is able to travel across multiple channels. However, most previous studies are limited to a single network or select multiple influential users based on the centrality ranking result of each user, not addressing the overlapping influence (OI) among users. In practice, the collective influence of multiple users is not equal to the total sum of these users’ influences. In this paper, we propose a novel OI-based method for identifying multiple influential users in multiplex social networks. We first define the effective spreading shortest path (ESSP) by utilizing the concept of spreading rate in order to denote the relative location of users. Then, the collective influence is quantified by taking the topological factor and the location distribution of users into account. The identified users based on our proposed method are central and relatively scattered with a low overlapping influence. With the Susceptible-Infected-Recovered (SIR) model, we estimate our proposed method with other benchmark algorithms. Experimental results in both synthetic and real-world networks verify that our proposed method has a better performance in terms of the spreading efficiency.
Introduction
The development of online social networks (OSNs) has created a new major interaction medium and formed promising landscape for information dissemination. The engagement of online users generates a huge volume of data for investigating the human behavioral patterns [1]. More importantly, the fact that an opinion or decision of individuals is influenced by their neighbors or friends has a considerable impact on the popularity of new products or brands [2], [3]. Targeting influential users is vital for designing techniques for either accelerating the information diffusion in marketing applications or suppressing the propagation of unwanted contents [4], [5]. The crucial problem is how to select multiple users, called central users, who can influence a massive number of users [6]. The measurement of influential users is beneficial for advertisers to implement effective campaigns. Central users are believed to play a key role in the propagation process. In practice, if a virus attacks a central user with a large degree, betweenness, PageRank or k-shell [7], [8], it would quickly pervade the whole network [9]. If we protect or immunize these users, the propagation scale would be greatly alleviated [10]. Although the propagation dynamics have received more and more attention, most of the studies still remain in a single network [11]. However, in fact, a user often has more than one social account such as Twitter, Facebook and Instagram.