SP-GCRL: Influence Maximization on Incomplete Social Graphs

The article presents SP-GCRL, a novel framework for influence maximization on incomplete social graphs, leveraging a social-propagation-aware nonlinear diffusion function to address challenges posed by noisy data and non-stationary dynamics. It employs dual structural views for contrastive learning to enhance node representation resilience against missing edges, and integrates a GAT-based regression surrogate to optimize efficiency. Experimental results demonstrate that SP-GCRL outperforms both heuristic and learning-based methods on various real-world networks, making it a significant advancement for practitioners focused on scalable influence maximization strategies in AI applications.