Dynamic Gradient Influencing for Viral Marketing Using Graph Neural Networks

Saurabh Sharma, Ambuj Singh

Keywords: Viral Marketing, Social Network Analysis, Influence Propagation

Abstract: The problem of maximizing the adoption of a product through the process of viral marketing in social networks is of extreme importance. Accordingly, we use Graph Neural Networks (GNN) to model the adoption of products in a data-driven way using both topological and attribute information. We propose the novel Dynamic Viral Marketing (DVM) problem of finding the minimum budget and minimal set of dynamic topological and attribute changes to attain a spread goal. We show that DVM is NP-Hard and is connected to influence maximization (IM). Motivated by this connection, we develop the Dynamic Gradient Influencing (DGI) model, which uses gradient-guided node flipping to search for the optimal perturbations and targets low-budget and high influence non-adopters in discrete steps. We use an efficient strategy for computing node budgets and develop the Meta-Influence heuristic for assessing a node’s downstream influence. We evaluate our proposed DGI against multiple non-gradient and gradient baselines to show the efficacy of our approach on real-world attributed networks. Experiments reveal that DGI discovers realistic cascade patterns through intermediary low-degree nodes that are confidently classified as adopters by the GNN.

Submission Type: Extended abstract (max 4 main pages).

Submission Number: 144