物理学报2016，Vol.65Issue(16)：168901-1-168901-12,12.DOI:10.7498/aps.65.168901
一种有效的基于三角结构的复杂网络节点影响力度量模型∗
An efficient no de influence metric based on triangle in complex networks
摘要
Abstract
Influential nodes in large-scale complex networks are very important for accelerating information propagation, understanding hierarchical community structure and controlling rumors spreading. Classic centralities such as degree, betweenness and closeness, can be used to measure the node influence. Other systemic metrics, such as k-shell and H-index, take network structure into account to identify influential nodes. However, these methods suffer some drawbacks. For example, betweenness is an effective index to identify influential nodes. However, computing betweenness is a high time complexity task and some nodes with high degree are not highly influential nodes. Presented in this paper is a simple and effective node influence measure index model based on a triangular structure between a node and its neighbor nodes (local triangle centrality (LTC)). The model considers not only the triangle structure between nodes, but also the degree of the surrounding neighbor nodes. However, in complex networks the numbers of triangles for a pair of nodes are extremely unbalanced, a sigmoid function is introduced to bound the number of triangles for each pair of nodes between 0 and 1. The LTC model is very flexible and can be used to measure the node influence on weighted complex networks. We detailedly compare the influential nodes produced by different approaches in Karata network. Results show that LTC can effectively identify the influential nodes. Comprehensive experiments are conducted based on six real complex networks with different network scales. We select highly influential nodes produced by five benchmark approaches and LTC model to run spreading processes by the SIR model, thus we can evaluate the efficacies of different approaches. The experimental results of the SIR model show that LTC metric can more accurately identify highly influential nodes in most real complex networks than other indicators. We also conduct network robustness experiment on four selected networks by computing the ratio of nodes in giant component to remaining nodes after removing highly influential nodes. The experimental results also show that LTC model outperforms other methods.关键词
复杂网络/节点影响力/三角结构/关键节点Key words
complex network/node influence/triangle/key node引用本文复制引用
韩忠明,陈炎,李梦琪,刘雯,杨伟杰..一种有效的基于三角结构的复杂网络节点影响力度量模型∗[J].物理学报,2016,65(16):168901-1-168901-12,12.基金项目
国家自然科学基金(批准号:61170112)、教育部人文社会科学研究基金项目(批准号:13YJC860006)和北京市教委科学研究面上项目(批准号：KM201410011005)资助的课题.* Project supported by the National Natural Science Foundation of China (Grant No.61170112), the Research Fund Project of the Ministry of Education of Humanities and Social Science, China (Grant No.13YJC860006), and the Scientific Research Common Program of Beijing Municipal Commission of Education, China (Grant No. KM201410011005) （批准号:61170112）
