search for




 

A statistical social network model for multi-level labeling nodes
Journal of the Korean Data & Information Science Society 2018;29:49-58
Published online January 31, 2018
© 2018 Korean Data and Information Science Society.

Jinkwang Kim1 · Changhyuck Oh2

12Department of Statistics, Yeungnam University
Correspondence to: Professor, Department of Statistics, Yeungnam University, Gyeongsan, Gyeongbuk 38541, Korea. E-mail: choh@yu.ac.kr
Received December 4, 2017; Revised January 6, 2018; Accepted January 9, 2018.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
In a social network, it is meaningful to estimate the citations between the nodes based on informations whether they have labels on some characteristics or not. In this paper, we consider a directional network model of which nodes have been attached with labels with multi-level nominal information, and under the assumption that nodes with similar levels on labels have high possibility to be linked each other, we assumed a regression model between levels of labels and reciprocation strength between node pairs. We assume a conditional model in which nodes not linked to any other nodes are excluded from the network considered and derive a likelihood function for the parameters of the model to predict the citations among the nodes whith multi-level information. The Monte Carlo simulation showed that the estimation of the parameters and the estimation of the social network using this method are valid. In a network with more nodes and less density, the proposed model showed better efficiency and the AUC tended to increase as more labels were included in the model.
Keywords : AUC, label similarity, likelihood function, multi-level label, stochastic network model.