Topic and opinion infused hypergraphs for influence maximization

Abstract

The modeling of online social network interactions is more efficiently achieved using hypergraphs, where hyperedges represent topics of user engagement. This work introduces a topic and opinion-infused hypergraph, where user opinion magnitude and polarity toward specific topics are incorporated into the incidence matrix. For a target topic, relevant hyperedges identify vulnerable targets, and a Topical Relevance Infused Incidence Matrix (TRIIV) is extracted to capture the participatory opinion of vulnerable users. Two novel seed selection criteria, based on participation and opinion severity with weighted usage, are proposed. Infection probability is set to 1 or determined randomly based on the seed's opinion similarity with vulnerable users. Experiments show that deterministic infection of a subset enhances influence spread efficiency while reducing computational requirements.