Maximum likelihood-based influence maximization in social networks
- PDF / 1,718,539 Bytes
- 16 Pages / 595.276 x 790.866 pts Page_size
- 67 Downloads / 268 Views
Maximum likelihood-based influence maximization in social networks Wei Liu 1,2 & Yun Li 1 & Xin Chen 1 & Jie He 1
# Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Influence Maximization (IM) is an important issue in network analyzing which widely occurs in social networks. The IM problem aims to detect the top-k influential seed nodes that can maximize the influence spread. Although a lot of studies have been performed, a novel algorithm with a better balance between time-consumption and guaranteed performance is still needed. In this work, we present a novel algorithm called MLIM for the IM problem, which adopts maximum likelihood-based scheme under the Independent Cascade(IC) model. We construct thumbnails of the social network and calculate the L-value for each vertex using the maximum likelihood criterion. A greedy algorithm is proposed to sequentially choose the seeds with the smallest L-value. Empirical results on real-world networks have proved that the proposed method can provide a wider influence spreading while obtaining lower time consumption. Keywords Influence Maximization . Independent Cascade Model . Maximum Likelihood
1 Introduction With the rapid development of the internet, social media which has increased connections through social networks [1] has flourished. A social network is comprised of social actors and their relationships. Researchers currently use graphs to represent social networks with nodes representing the social individuals and edges representing their relations. In the real world, this relationship can be competitors, collaborators, hostiles, friends, etc. In recent years, following the rise of social networks like Facebook, Youtube, Whatsapp, LinkedIn, and Twitter, a lot of studies based on social networks have drawn research attention [2, 3]. It is very interesting and important to establish appropriate models to effectively disseminate information in social networks. From a marketing strategy perspective, such as “viral marketing” and “word of mouth” [4–8], marketers can select
* Yun Li [email protected] 1
College of Information Engineering of Yangzhou University, Yangzhou 225127, China
2
The Laboratory for Internet of Things and Mobile Internet Technology of Jiangsu Province, Huaiyin Institute of Technology, Huaiyin 223002, China
representative individual subsets in the population to make them generate larger cascading effects [9, 10]. The above problem is called the influence maximization problem, which has now become a hot research issue. Domingos et al. [5] introduced the IM problem with a mathematical model. Subsequently, a variety of spreading models and methods have been addressed for the problem. A precise understanding of influence over social networks is indispensable in advertising promotion, political elections, community formation, and so on. There are two challenges to be solved: (1) design an appropriate model which could properly simulate the diffusion of influence in social networks, (2) develop an efficient approach to determ
Data Loading...
