VISUAL ANALYSIS OF HETEROGENEOUS AND DYNAMIC GRAPHS

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence


Title: "VISUAL ANALYSIS OF HETEROGENEOUS AND DYNAMIC GRAPHS"

By

Miss Panpan Xu


Abstract

Many real world problems can be modeled as heterogeneous graphs where 
nodes and/or links are of different types. And these graphs are often 
dynamically changing. One example is the bibliographic database, where 
authors and research topics are different types of entities, and the graph 
changes over time as the researchers switch their interests or form new 
collaborative relations. Research in the area of graph visualization has 
been concerned with designing novel and effective visual encoding schemes 
and user interactions for the viewers to gain insight into graph data. We 
follow this line of research and this thesis reports our work in 
developing visual analysis techniques for heterogeneous and dynamic graph 
data from various application domains.

In the first work, we visualize heterogeneous graph data that not only 
records the relationship among people, but also the various items they are 
related to (e.g. interested topics or music). We design visualizations 
that can help to study if people closely linked have similar items of 
interest, and introduce a novel set visualization technique and the 
corresponding layout algorithm to display the overlap of their interests. 
The techniques are applied to a bibliographic dataset and the user data 
from a social music service website.

The second work studies the dynamic interplay among topics, opinion 
leaders and the audiences on social media. More specifically, we propose a 
combination of time series modeling and interactive visualization 
techniques to study how various topics compete to attract public attention 
when they are spreading on social media (e.g. Twitter), and what roles do 
opinion leaders such as mass media, political figures and grassroots play 
in the rise and fall of various topics. In the experiment, we report the 
insights gained on collections of Tweets.

The third study proposes a visualization technique to explore network 
dynamics, especially how the new edges are formed through the assortative 
and relational mechanisms, which have been observed in the evolution of 
many networks. The visualization technique developed not only displays the 
structural evolution of a dynamic network, but also allows the viewer to 
explore the various mechanisms underlying the changes. The techniques are 
demonstrated through the visual analysis of real-world datasets: the 
co-authorship network and the user interaction graph on social websites.


Date:			Monday, 15 December 2014

Time:			3:00pm - 5:00pm

Venue:			Room 3501
 			Lifts 25/26

Chairman:		Prof. Shengwang Du (PHYS)

Committee Members:	Prof. Huamin Qu (Supervisor)
 			Prof. Pedro Sander
 			Prof. Chiew-Lan Tai
 			Prof. Yongshun Cai (SOSC)
                        Prof. Xiaoru Yuan (EECS, Peking Univ.)
 			Prof. Wenjie Li (Computing, PolyU)


**** ALL are Welcome ****