An influence model for influence maximization–revenue optimization

Abstract

The rise of online social networks (OSNs) has caused an insurmountable amount of interest from advertisers and researchers seeking to monopolize on its features. Researchers aim to develop strategies for propagating information among users within an OSN that is captured by diffusion or influence models. Within the last decade, influence models have been extensively studied for the influence maximization (IM) problem. Recently, a novel stochastic dynamic programming (SDP) formulation to influence maximization called the influence maximization–revenue optimization (IM–RO) problem was proposed with numerous lucrative advantages. In this paper, we validate the intuition behind the proposed influence model for the IM–RO problem empirically. We focus on demonstrating the correctness of the notion behind the influence model that the more of a user’s friends who click on an advertisement, the more likely the user is to click on the advertisement themselves and use a decision tree regressor to predict this probability. To further support the premise of our influence model and estimate its parameters, we implement a linear regression and a Bayesian model. Results indicate that the linear regression model captures the functional relationship between its dependent and the independent variables. We extend the experiments to real-world OSNs and investigate additional predictor variables that influence the number of posts and reposts.