Mining set of influencers in signed social networks with maximal collective influential power: A genetic algorithm approach

Abstract

The ubiquitous growth of social networks opens a new line of research for developing algorithms and models for influence mining. Determining influential people in the network which consists of both positive and negative links between users is a challenging task. It becomes critical for businesses with fixed budget constraints to identify a group of influential people whose views will influence others’ behaviors the most. In this paper, we propose a model that aims to discover a set of influencers in signed social networks with maximal Collective Influential Power (CIP). We first construct an “influence network” between users and compute the influence strength between each pair of users by utilizing both the explicit trust-distrust information provided by users and the information derived from interactions between them. We then employ an elitist genetic algorithm that discovers a set of influencers with high influence spread as well as maximal enhanced joint influential power over the other users in the network. Experiments are performed on Epinions, a real-world dataset, and the results obtained are quite promising and clearly demonstrate the effectiveness of our proposed model.