Influence Maximization With Deactivation In Social Networks

TL;DR

This paper extends the influence maximization problem in social networks to a competitive setting where a leader aims to maximize influence and a follower seeks to minimize it through deactivation, using bilevel modeling and heuristics.

Contribution

It introduces a bilevel model for competitive influence maximization with deactivation, solved via enumeration and matheuristics, incorporating stochastic approximation methods.

Findings

Effective heuristic solutions for large instances.

Bilevel model captures competitive influence dynamics.

Approximation techniques improve solution scalability.

Abstract

In this paper we consider an extension of the well-known Influence Maximization Problem in a social network which deals with finding a set of k nodes to initiate a diffusion process so that the total number of influenced nodes at the end of the process is maximized. The extension focuses on a competitive variant where two decision makers are involved. The first one, the leader, tries to maximize the total influence spread by selecting the most influential nodes and the second one, the follower, tries to minimize it by deactivating some of these nodes. The formulated bilevel model is solved by complete enumeration for small-sized instances and by a matheuristic for large-sized instances. In both cases, the lower level problem, which is a stochastic optimization problem, is approximated via the Sample Average Approximation method.