基于传播概率熵的重要节点识别方法

Identifying Vital Nodes Based on Epidemic Spreading Probability Entropy

摘要: 本研究提出了一种新的方法ESPE (传播概率熵)，旨在更有效地评估复杂网络中节点的重要性。该方法通过结合信息熵同时考虑节点自身的传播影响力及其邻居节点的传播影响，解决了以往研究中忽略邻居节点影响力的问题。为了验证ESPE算法的性能，采用肯德尔相关系数来衡量其识别节点传播能力的准确性，通过与SIR(易感–感染–恢复)传播模型的比较，证明了ESPE在识别关键节点方面的优越性。实验结果表明，ESPE算法在六个真实网络中平均提高了8.46%，其Kendall相关系数普遍超过现有方法。此外，在分析ESPE和SPC算法获得的归一化传播影响力分数之间的相关性时，ESPE比SPC表现出更高的准确性和判别力。总体而言，ESPE算法为复杂网络分析提供了准确可靠的工具，能够更有效地识别和评估网络中的关键节点。

Abstract: This study introduces a novel method known as ESPE (Epidemic Spreading Probability Entropy), designed to more effectively evaluate the importance of nodes within complex networks. The method addresses the oversight in previous research, which neglected the influence of neighboring nodes, by incorporating information entropy to simultaneously consider a node’s own spreading influence and the spreading impact of its neighbors. To validate the performance of the ESPE algorithm, Kendall’s tau rank correlation coefficient was employed to measure its accuracy in identifying the spreading capabilities of nodes. The superiority of ESPE in identifying key nodes was demonstrated through comparison with the SIR (Susceptible-Infected-Recovered) spreading model. Experimental results indicate that the ESPE algorithm achieved an average improvement rate of 8.46% across six real-world networks, with its Kendall correlation coefficients generally exceeding those of existing methods. Furthermore, in the analysis of the correlation between the normalized spreading influence scores obtained by the ESPE and SPC algorithms, ESPE demonstrated greater accuracy and discriminative power compared to SPC. Overall, the ESPE algorithm provides an accurate and reliable tool for complex network analysis, capable of more effectively identifying and assessing key nodes within the network.