Scholarlink

<p>Identifying influential spreaders in social networks is essential for optimizing information diffusion and controlling epidemic outbreaks. Traditional centrality-based methods, typically applied to undirected and unweighted networks, often fail to capture the complex dynamics of real-world weighted directed networks. To address this gap, we introduce a novel approach, the Weighted Crossbred Method-based Graph Convolutional Network (WCM-GCN). This model integrates both structural and dynamic attributes of weighted directed networks to more effectively identify key spreaders. The WCM-GCN leverages two crucial features that measure a node’s weighted spreading capacity to spread information throughout the network. These features are embedded into a weighted crossbred framework within a GCN architecture, enabling the model to learn both topological influence and dynamic propagation ability. Unlike traditional GCNs, WCM-GCN mitigates over-smoothing and avoids the information loss commonly introduced by DropEdge regularization. We evaluate WCM-GCN across twelve real-world weighted directed networks, comparing its performance against eleven established baseline methods. Experimental results using the WDSIR epidemic model show that WCM-GCN consistently achieves superior influence spread, outperforming both the original Weighted Crossbred Method and various centrality-based approaches. Thus, WCM-GCN presents a robust and scalable solution for identifying influential spreaders in complex networks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

WCM-GCN: identifying influential spreaders using graph convolutional networks on weighted directed networks based on spreading properties

  • 期刊论文

  • 发表日期：

    2026年1月4日

查看全文
  • Nilanjana Saha,
  • Animesh Dutta

摘要

Identifying influential spreaders in social networks is essential for optimizing information diffusion and controlling epidemic outbreaks. Traditional centrality-based methods, typically applied to undirected and unweighted networks, often fail to capture the complex dynamics of real-world weighted directed networks. To address this gap, we introduce a novel approach, the Weighted Crossbred Method-based Graph Convolutional Network (WCM-GCN). This model integrates both structural and dynamic attributes of weighted directed networks to more effectively identify key spreaders. The WCM-GCN leverages two crucial features that measure a node’s weighted spreading capacity to spread information throughout the network. These features are embedded into a weighted crossbred framework within a GCN architecture, enabling the model to learn both topological influence and dynamic propagation ability. Unlike traditional GCNs, WCM-GCN mitigates over-smoothing and avoids the information loss commonly introduced by DropEdge regularization. We evaluate WCM-GCN across twelve real-world weighted directed networks, comparing its performance against eleven established baseline methods. Experimental results using the WDSIR epidemic model show that WCM-GCN consistently achieves superior influence spread, outperforming both the original Weighted Crossbred Method and various centrality-based approaches. Thus, WCM-GCN presents a robust and scalable solution for identifying influential spreaders in complex networks.