如何定量评估复杂网络中节点的影响力是一个重要的研究课题，因为它有助于深入理解网络的结构和功能。现有的多数方法主要基于网络固有拓扑的分析建立，缺少对多阶邻居节点的传播性质和拓扑信息的综合利用，然而它们对影响力节点识别有重要影响。为此，本文提出了一种综合多阶邻居传播度量和拓扑特征(multi-order neighbor propagation metrics and topological features，MNPMTF)的算法来有效识别复杂网络中的影响力节点。首先，该算法结合传播模型和最短路径来刻画邻居节点的传播概率，从而量化节点之间信息传播的可能性。其次，考虑多阶邻居中的邻居重叠比形成邻居重叠度，进而量化信息在邻居网络中的传播路径。再次，利用节点的k壳、h指数和聚类系数构成新指标KHC系数，以此来描述节点的拓扑特征。最后，算法综合3阶邻居范围内的传播概率、邻居重叠度和拓扑特征以评估节点的影响力。在9个真实网络上的大量实验表明，所提算法在排序准确性、有效性和区分能力等多方面均优于7种具有代表性的方法，为复杂网络中节点影响力评估提供了一种新的思路。
Identification of high-influential nodes based on multi-order neighbor propagation metrics and topological features
How to quantitatively evaluate the influence of nodes in complex networks is an important research topic because it helps to understand the structure and functionality of networks.Most existing methods are primarily based on analyzing the network's inherent topology,needing more comprehensive utilization of propagation properties of multi-order neighboring nodes and topological information,significantly impacting identifying influential nodes.To this end,this paper proposes an algorithm called multi-order neighbor propagation metrics and topological features(MNPMTF)to effectively identify influential nodes in complex networks.Firstly,the algorithm combines a propagation model and shortest paths to characterize the propagation probability of neighboring nodes,thereby quantifying the likelihood of information spreading between nodes.Secondly,the algorithm considers the neighbor overlap ratio among multi-order neighbors to form the neighbor overlap degree,thereby quantifying the propagation paths of information in the neighbor network.Thirdly,the algorithm utilizes the k-shell index,h-index,and clustering coefficient to form a new index called the KHC coefficient,which describes the topological features of nodes.Finally,the algorithm integrates the propagation probability,neighbor overlap degree,and topological features within a 3-order neighbor range to evaluate the influence of nodes.Extensive experiments on nine entire networks demonstrate that the proposed algorithm outperforms seven representative methods regarding ranking accuracy,effectiveness,and discriminability,providing a new approach for assessing node influence in complex networks.
NETL
NSTL
万方数据
