基于高阶反向影响采样的无向超图影响力最大化算法OA

Existing studies on undirected hypergraph influence maximization generally lack in-depth analysis and effec-tive modeling of hypergraph higher-order structures,making it difficult to characterize the complex group-level informa-tion diffusion mechanisms.Meanwhile,most existing algorithms rely heavily on local topological features and fail to capture collective behaviors.To address these issues,information diffusion in undirected hypergraphs was investigated systematically,two typical group-based information diffusion processes were formalized,and an undirected hypergraph independent cascade model was proposed.Based on this model and the classic reverse influence sampling framework,a hypergraph influence maximization algorithm based on higher-order reverse influence sampling was developed,and is was theoretically proved that it achieved a(1-1/e-ε)approximation to the optimal solution.Experiments on eight real-world hypergraph datasets show that the proposed method significantly outperforms all baseline approaches,improving the influence spread by up to 42.85%.Meanwhile,its average runtime is only 0.7%of the CELF algorithm and is compa-rable to heuristic-based methods.