Dynamic Graph Embedding for Social Network Influence Propagation

Abstract

Social network influence propagation is a complex phenomenon significantly impacted by the dynamic nature of user interactions and evolving network structures. Traditional static graph embedding methods, which primarily focus on capturing the structural information of a network at a single point in time, often fall short of accurately predicting viral trends and understanding how influence spreads over time. This paper introduces a novel dynamic graph embedding approach designed specifically to address this limitation. Our method leverages recurrent neural networks (RNNs) to model the temporal dependencies inherent in influence spread, continuously updating node embeddings based on both the evolving network topology and the sequence of user interactions. This dynamic embedding strategy allows for a more nuanced representation of nodes, reflecting their changing influence potential within the network. We demonstrate through theoretical analysis and a conceptual framework that our approach offers a significant improvement over static methods in capturing the temporal dynamics crucial for accurately predicting influence propagation patterns. The core contribution lies in the integration of temporal modeling with graph embedding, enabling a richer understanding of social influence propagation.