Scholarlink

<p>Temporal graphs model the dynamic evolution of nodes and edges and are widely used in domains such as social networks and academic collaboration. The (<i>k</i>,&#xa0;<i>h</i>)-core is a temporal subgraph model where each node connects to at least <i>k</i> others, and each node pair shares at least <i>h</i> temporal edges, reflecting <i>interaction strength</i>. Existing studies focus on (<i>k</i>,&#xa0;<i>h</i>)-core decomposition and maintenance but lack efficient query solutions. To fill this gap, we propose two baselines: an online method and a Full Index method. The online method avoids index construction but offers low query efficiency; the Full Index supports fast queries but incurs high storage cost. To balance efficiency and index size, we design a Tree Index based on node set partitioning and an enhanced Permutation Tree Index. Both significantly reduce storage overhead while supporting fast queries. Experiments on real-world datasets show that our indexes are more compact than the Full Index and consistently deliver better query performance than both baselines.</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

Mining Interaction-intensive cores in temporal graphs

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月10日

查看全文
  • Luo Zhou,
  • Xueying Gong,
  • Jinhu Lu,
  • Yue Liang,
  • Huixiang Wen

摘要

Temporal graphs model the dynamic evolution of nodes and edges and are widely used in domains such as social networks and academic collaboration. The (k, h)-core is a temporal subgraph model where each node connects to at least k others, and each node pair shares at least h temporal edges, reflecting interaction strength. Existing studies focus on (k, h)-core decomposition and maintenance but lack efficient query solutions. To fill this gap, we propose two baselines: an online method and a Full Index method. The online method avoids index construction but offers low query efficiency; the Full Index supports fast queries but incurs high storage cost. To balance efficiency and index size, we design a Tree Index based on node set partitioning and an enhanced Permutation Tree Index. Both significantly reduce storage overhead while supporting fast queries. Experiments on real-world datasets show that our indexes are more compact than the Full Index and consistently deliver better query performance than both baselines.