Scholarlink

Link prediction is a vital component for the optimization and management of telecommunications networks. However, understanding the structure, dynamics, and hidden communication patterns within phone call networks poses a significant challenge, due to the complexity of user interactions and the lack of predefined information regarding their relationships. This paper explores the application of graph neural networks (GNNs) in link prediction, focusing on two prominent models: Graph Convolutional Network (GCN) and Graph Attention Network (GAT). By utilizing call detail records (CDRs), a graphical representation of the network is constructed, and the performance of these models in predicting future links is evaluated. Our results demonstrate that the GCN model achieves a maximum accuracy of 0.8636, while the GAT attains an accuracy of 0.4943. Furthermore, the impact of community structure on link prediction performance is examined, suggesting that incorporating community information could enhance model accuracy.

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

Link Prediction in Telecommunications Networks Using Graph Neural Networks and Community Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Valentina Sanhueza-Campos,
  • Gustavo Gatica,
  • Alex Barrales-Araneda,
  • Jairo R. Coronado-Hernandez

摘要

Link prediction is a vital component for the optimization and management of telecommunications networks. However, understanding the structure, dynamics, and hidden communication patterns within phone call networks poses a significant challenge, due to the complexity of user interactions and the lack of predefined information regarding their relationships. This paper explores the application of graph neural networks (GNNs) in link prediction, focusing on two prominent models: Graph Convolutional Network (GCN) and Graph Attention Network (GAT). By utilizing call detail records (CDRs), a graphical representation of the network is constructed, and the performance of these models in predicting future links is evaluated. Our results demonstrate that the GCN model achieves a maximum accuracy of 0.8636, while the GAT attains an accuracy of 0.4943. Furthermore, the impact of community structure on link prediction performance is examined, suggesting that incorporating community information could enhance model accuracy.