GEHIM: A Graph Embedding-Based Heterogeneous Influence Maximization Algorithm

Influence maximization in heterogeneous information networks, which contain multiple types of nodes and relationships, has significant applications in marketing strategies and information propagation analysis. The objective is to identify a minimal set of seed nodes that can maximize the influence spread across the network. A key challenge in this domain lies in the complex and diverse influence propagation patterns between different types of nodes, particularly in effectively integrating heterogeneous node information and semantic relationships to accurately model the influence propagation process. While existing metapath-based approaches consider semantic connections between nodes, they still fall short in modeling these complex propagation patterns. To address this challenge, this paper proposes Graph Embedding-based Heterogeneous Influence Maximization (GEHIM), a novel algorithm that effectively enhances the fusion of both intra-metapath and cross-metapath information through attention mechanisms. Notably, experimental results on two heterogeneous information networks demonstrate that GEHIM improves the influence spread by [Formula: see text] under the linear threshold model and by [Formula: see text] under the independent cascade model, compared to the state-of-the-art baselines.