社交网络中通过添加链接提升影响力排名
Influence Ranking Improvement via Link Addition in Social Networks
- Graduate School of Science and Technology, University of Tsukuba（筑波大学科学技术研究院）
- Institute of Systems and Information Engineering, University of Tsukuba（筑波大学系统信息工程学院）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
本研究探讨通过添加少量出链提升社交网络中目标节点的影响力排名，提出贪心和随机搜索两种策略，实验表明少量链接即可显著提升排名，并揭示了排名提升与计算效率的权衡。
AI中文摘要：
社交媒体平台越来越依赖有影响力的用户进行信息传播，例如在营销和政治竞选等领域。随着被认定为有影响力的价值增长，用户可能有动机战略性地提升自身影响力。在本研究中，我们探讨在影响力最大化（IM）背景下，通过添加有限数量的出链，目标节点的影响力排名是否以及能在多大程度上得到提升。我们将影响力排名提升的链接选择问题（LSP-IRI）形式化为从目标节点选择固定大小的额外出链集合以最大化其排名提升的问题。为研究该问题，我们考虑了两种代表性的启发式策略：一种直接优化排名提升的贪心方法，以及一种计算高效的随机搜索方法。我们在四个真实网络（节点规模从数千到数十万不等）上进行了实验。结果表明，即使添加少量链接也能显著提升基于IM的影响力排名。特别是，贪心方法仅添加三条链接即可使初始排名约在第50位的节点平均提升数十个名次，有时甚至进入前10名。随机搜索方法在严格预算下获得的提升较小，但与贪心方法相比，计算时间最多减少约98%，并且在允许更大预算时变得有效。这些发现表明，基于IM的影响力排名对有限的局部结构修改敏感，并凸显了排名提升与计算效率之间的权衡。
英文摘要：
Social media platforms increasingly rely on influential users for information dissemination in domains such as marketing and political campaigns. As the value of being recognized as influential grows, users may have incentives to strategically enhance their influence. In this study, we investigate whether and to what extent the influence ranking of a target node can be improved by adding a limited number of outgoing links in the context of influence maximization (IM). We formulate the Link Selection Problem for Influence Ranking Improvement (LSP-IRI) as the problem of selecting a fixed-size set of additional outgoing links from a target node in order to maximize its rank improvement. To examine this problem, we consider two representative heuristic strategies: a greedy method that directly optimizes rank improvement and a computationally efficient random search method. We conduct experiments on four real-world networks ranging from thousands to hundreds of thousands of nodes. The results show that even a few added links can substantially improve IM-based influence rankings. In particular, the greedy method improves the ranks of nodes initially ranked around 50 by several tens of positions on average with only three added links, sometimes moving them into the top 10. The random search method achieves smaller gains under strict budgets but reduces computation time by up to approximately 98% compared with the greedy method and becomes effective when larger budgets are allowed. These findings show that IM-based influence rankings are sensitive to limited local structural modifications and highlight a trade-off between ranking improvement and computational efficiency.
