Revealing of the Underlying Mechanism of Different Node Centralities Based on Oscillation Dynamics on Networks

Chisa TAKANO  Masaki AIDA  

IEICE TRANSACTIONS on Communications   Vol.E101-B   No.8   pp.1820-1832
Publication Date: 2018/08/01
Online ISSN: 1745-1345
DOI: 10.1587/transcom.2017EBP3370
Type of Manuscript: PAPER
Category: Fundamental Theories for Communications
oscillation dynamics,  node centrality,  social network analysis,  

Full Text: PDF>>
Buy this Article

In recent years, with the rapid development of the Internet and cloud computing, an enormous amount of information is exchanged on various social networking services. In order to handle and maintain such a mountain of information properly by limited resources in the network, it is very important to comprehend the dynamics for propagation of information or activity on the social network. One of many indices used by social network analysis which investigates the network structure is “node centrality”. A common characteristic of conventional node centralities is that it depends on the topological structure of network and the value of node centrality does not change unless the topology changes. The network dynamics is generated by interaction between users whose strength is asymmetric in general. Network structure reflecting the asymmetric interaction between users is modeled by a directed graph, and it is described by an asymmetric matrix in matrix-based network model. In this paper, we showed an oscillation model for describing dynamics on networks generated from a certain kind of asymmetric interaction between nodes by using a symmetric matrix. Moreover, we propose a new extended index of well-known two node centralities based on the oscillation model. In addition, we show that the proposed index can describe various aspect of node centrality that considers not only the topological structure of the network, but also asymmetry of links, the distribution of source node of activity, and temporal evolution of activity propagation by properly assigning the weight of each link. The proposed model is regarded as the fundamental framework for different node centralities.