Mining Business Value from Dynamic and Multi-View Company Networks : A Latent Space Model Approach

Abstract

Numerous research has been devoted to the study of social network, while relatively little research is related to business network. Besides, the dynamic nature and heterogeneous types of edges in network still remain under-researched. In this paper, an advanced statistical network model is proposed and extended to analyze dynamic and multi-view company networks. Multi-view refers to heterogeneous tyeps of edges among same set of nodes. The statistical model assumes that the probability of link between a pair of nodes depends only on the underlying unobserved space. Therefore, nodes that are close to each other in the unobserved space are more likely to have link. The statistical model is inferenced within the Bayesian framework, and the parameters are estimated using Markov Chain Monte Carlo (MCMC) procedures. We demonstrate the empirical value of our model by applying it to two company networks, the investment network and the news network, with same set of nodes. The investment network is constructed from investment transactions collected from the Thomson Reuters Eikon while the news network is constructed from financial news collected from the Reuters site archive. We show that our model can be applied in many business applications such as measuring business proximity, studying business influence, understanding alliance structure and predicting business relationship.

Date of Award	20 Aug 2018
Original language	English
Awarding Institution	Hong Kong University of Science and Technology
Supervisor	Kar Yan Tam (Chief supervisor) & Mike Ka Pui So (Co-supervisor)