Ji Yating, Liu Lequn, Li Shujia, Lu Pu, Tang Qimei
Information Management Office, Hefei Normal University, Hefei, China.
Key Laboratory of Philosophy and Social Science of Anhui Province on Adolescent Mental, Health and Crisis Intelligence Intervention, Hefei Normal University, Hefei, China.
PLoS One. 2025 Jul 24;20(7):e0328381. doi: 10.1371/journal.pone.0328381. eCollection 2025.
Identifying key nodes in complex networks holds significant application value in fields such as information dissemination and disease spread. The traditional K-shell decomposition method has low time complexity and is suitable for large-scale complex networks; however, it only considers global positional information, leading to lower discrimination. To improve the K-shell decomposition method, many approaches have been proposed by researchers. However, there no algorithm has yet that simultaneously uses the iteration factor and degree to further distinguish nodes with the same K-shell value. To address this issue, we propose a node influence ranking algorithm that integrates K-shell iteration, node degree, and neighbor information, considering both global network position and local topology. Through simulation experiments on eight networks, it was verified that this method provides more accurate ranking results compared to dc, bc, cc, k-shell, Ks + , KSIF, LGI and DCK methods on eight networks, with an average accuracy improvement of 5.15% over the second-best algorithm. In identifying the top 10 key nodes, the KTD algorithm demonstrates higher accuracy than other methods. Additionally, it shows high discriminative power and good time performance, making it suitable for large-scale complex networks.
识别复杂网络中的关键节点在信息传播和疾病传播等领域具有重要的应用价值。传统的K-shell分解方法时间复杂度低，适用于大规模复杂网络；然而，它只考虑全局位置信息，导致区分度较低。为了改进K-shell分解方法，研究人员提出了许多方法。然而，目前还没有算法能够同时使用迭代因子和度来进一步区分具有相同K-shell值的节点。为了解决这个问题，我们提出了一种节点影响力排序算法，该算法整合了K-shell迭代、节点度和邻居信息，兼顾了全局网络位置和局部拓扑结构。通过对八个网络的模拟实验验证，该方法在八个网络上比dc、bc、cc、k-shell、Ks +、KSIF、LGI和DCK方法提供了更准确的排序结果，比次优算法的平均准确率提高了5.15%。在识别前10个关键节点时，KTD算法比其他方法具有更高的准确率。此外，它还具有较高的区分能力和良好的时间性能，适用于大规模复杂网络。
