DSGCN: A Degree Strength Graph Convolution Network for Identifying Influential Nodes in Complex Networks

Identifying influential nodes in complex networks is crucial for applications such as information control, viral marketing, and pricing strategy analysis. However, traditional centrality methods have limitations when applied to weighted networks with varying connection intensities. To address this challenge, we introduce Degree Strength (DS) based Graph Convolution Network (DSGCN). This GCN model integrates node degree and strength as features to accurately identify influential nodes in weighted networks. We conducted a performance analysis, comparing the effectiveness of DS with traditional centrality methods. Additionally, we evaluated the
$GCN$
-based approach that integrates DS and traditional centrality methods as features to assess its performance. The experimental analysis demonstrates that leveraging
$GCN$
on D S outperforms existing methods to capture ranking similarity and accurately identify influential spreaders in complex weighted networks. Evaluation metrics such as Kendall
$\tau$
correlation and improvement percentage consistently show the superiority of DSGCN. By effectively handling connected and disconnected networks, DSGCN provides valuable insights into understanding influence dynamics in real-world scenarios.