Evaluating user influence in social networks using k-core

Abstract

Given a social network with an influence propagation model, selecting a small subset of users to maximize the influence spread is known as influence maximization problem. It has been shown that influence maximization problem is NP-hard, and several approximation algorithms and heuristics have been proposed. In this work, we follow a graph-theoretic approach to find the initial spreaders called seed nodes such that the expected number of influenced users is maximized. It has been well established through a series of research works that a special subgraph called k-core is very useful to find most influential users. A k-core subgraph H of a graph G is defined as a maximal induced subgraph where every node in H is having at least k neighbors. We apply a topology-based algorithm called Local Index Rank (LIR) on k-core (for some fixed k) to select the seed nodes in a social network. The accuracy and efficiency of the proposed method have been established using two benchmark datasets of SNAP (Stanford Network Analysis Project) database.