Publication Type

PhD Dissertation

Version

publishedVersion

Publication Date

2-2014

Abstract

This dissertation addresses the subject of measuring social correlation among users within a complex social network. Social correlation is closely related to the measurement of social influence in social sciences. While social influence focuses on the existence of causal influence among users, we take a computational approach to measure correlation strength among users based on their shared interactions. We call this social correlation. To formally model social correlation, we propose a framework which contains two major parts. The first part is that of representing users behavior in a computationally efficient and accurate manner. For example, social media users perform many kinds of actions online such as buying products, watching videos and posting comments. The huge number of users’ actions logged over long duration poses significant challenges for analysis. We propose both static and temporal models to compress the huge amounts of users’ action data into low dimensional representations. For the dynamic users’ action data, there is the additional challenge of temporal sparsity where users have low amounts of activities in some time periods. This results in the lack of information in some time periods for modeling the temporal behavior of users. By exploiting the transition of users behavior in different time periods, we obtained a smoothed representation of users behavior in low dimensions. The second part of modeling social correlation is to take the users’ behavior in low dimensional representation and compare against the behaviors of other users whom they had earlier interacted with. The dissertation first proposes social correlation measurement for the static representation of users’ behavior. It then extends the measurement to the temporal case by using only two time periods and finally for the general case of multiple time periods using Granger causality. With our proposed set of social correlation measurements one can now build better recommendation systems that predict the missing or future users’ behavior considering the influence among users in the complex networks.

Keywords

social, networks, influence, latent, spaces, correlation

Degree Awarded

PhD in Information Systems

Discipline

Databases and Information Systems | Numerical Analysis and Scientific Computing | Social Media

Supervisor(s)

LIM, Ee-Peng

First Page

1

Last Page

179

Publisher

Singapore Management University

City or Country

Singapore

Copyright Owner and License

Author

Download

Share

COinS