Scholarlink

Graph clustering plays a crucial role in graph data analysis with applications in social networks, recommender systems, and bioinformatics. Traditional clustering methods often rely on static assumptions and fixed graph structures, which limit their performance in handling high-dimensional and sparse graph data. In this study, we propose a novel method that integrates graph embedding via Graph Convolutional Networks (GCNs) with an adaptive clustering algorithm. The GCN-based embedding captures both local and global structural information, while the adaptive clustering strategy dynamically adjusts parameters based on local graph characteristics. Experimental results on benchmark datasets demonstrate that our approach achieves superior accuracy, stability, and robustness compared to traditional spectral clustering and existing GCN-based methods, particularly in complex graph environments.

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

A Graph Clustering Method Based on Graph Embedding and Adaptive Clustering Strategies

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mengfei Li,
  • Jiayi Huang,
  • Hanyi Chen,
  • Xiangxin Xiao,
  • Haotian Wang,
  • Guangzhen Yao,
  • Wenxin Zhang,
  • Tianyu Hu,
  • Renda Han,
  • Fa Fu

摘要

Graph clustering plays a crucial role in graph data analysis with applications in social networks, recommender systems, and bioinformatics. Traditional clustering methods often rely on static assumptions and fixed graph structures, which limit their performance in handling high-dimensional and sparse graph data. In this study, we propose a novel method that integrates graph embedding via Graph Convolutional Networks (GCNs) with an adaptive clustering algorithm. The GCN-based embedding captures both local and global structural information, while the adaptive clustering strategy dynamically adjusts parameters based on local graph characteristics. Experimental results on benchmark datasets demonstrate that our approach achieves superior accuracy, stability, and robustness compared to traditional spectral clustering and existing GCN-based methods, particularly in complex graph environments.