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

PhD Thesis Defence



Mr. Wenchao WU


Without doubt, we are in the midst of a data explosion. A variety of data 
tracking human mobility, namely human mobility data, has been generated and 
collected within urban context, providing unprecedented opportunities to 
understand regional dynamics in urban area, which is of great social and 
business value in a variety of applications. However, due to large volumes and 
dynamic correlations of these data as well as high complexity of analytical 
tasks in real world applications, it is challenging for analysts to carry out 
in-depth analysis and extract valuable information. It often requires 
integrating human perception in the data exploration process, triggering a 
broad use of visual analytics. With visual analytics, we can include human 
perception in the data exploration process efficiently and combine the 
flexibility, creativity and domain knowledge of human beings with enormous 
storage capacity and computational power of today’s computers.

In this thesis, we introduce three advanced visual analysis techniques for 
uncovering regional dynamics in urban area from different aspects based on 
heterogeneous human mobility data. In particular, we first study the subject 
matter of regional boundary change and present BoundarySeer. It is a visual 
analytics system consisting of four major viewers to facilitate the general 
analytical tasks dealing with boundary changes of a region in urban area. 
Secondly, a visual analytics system, TelCoVis, is presented to facilitate the 
exploration of co-occurrence in human mobility (i.e. people from two regions 
visit an urban place during the same time span) and hidden correlations based 
on telco data. The system integrates a novel contour-based treemap with 
extended visualization techniques to enhance analysts’ perception for a 
comprehensive exploration of coordinated relationships among different regions 
and identify interesting patterns. The third study proposes a novel visual 
analysis approach to investigate people’s activity patterns for an interactive 
region segmentation based on three types of heterogeneous mobility data (i.e. 
taxi trajectories, metro passenger RFID card records and telco data). Combining 
advanced visualization techniques (e.g. NMF-based method to capture latent 
activity patterns, as well as metric learning to calibrate and supervise the 
underlying analysis) with intuitive visual designs (e.g. a voronoi-based 
texture map with elliptical activity glyphs to summarize people’s activities 
and enable a fast comparison), MobiSeg not only makes it easier for domain 
experts to perform a series of analyses on region segmentation, but also 
enables a new way to explore data from multiple levels and perspectives.

To the best of our knowledge, the above techniques are cutting-edge studies of 
visually analyzing regional dynamics in urban area based on heterogeneous human 
mobility data. To validate the effectiveness and usefulness of our study, all 
the proposed techniques and systems are deployed to analyze real-world datasets 
and evaluated by domain experts or target users.

Date:			Friday, 12 August 2016

Time:			1:00pm - 3:00pm

Venue:			Room 5564
 			Lifts 27/28

Chairman:		Prof. Jidong Zhao (CIVL)

Committee Members:	Prof. Lionel Ni (Supervisor)
 			Prof. Huamin Qu (Supervisor)
 			Prof. Cunsheng Ding
 			Prof. Xiaojuan Ma
 			Prof. Zongjin Li (CIVL)
 			Prof. Xiaohua Jia (CityU)

**** ALL are Welcome ****