Combining Automated Analysis with Interactive Visualizations for Spatio-temporal Data Analysis

PhD Thesis Proposal Defence


Title: "Combining Automated Analysis with Interactive Visualizations for 
Spatio-temporal Data Analysis"

by

Mr. Dongyu LIU


Abstract:

The rapid advances in sensing technologies and large-scale computing 
infrastructures lead to explosive growth in data. Spatio-temporal data, as 
one of the ubiquitous data types, is increasingly collected and 
extensively studied in various scientific domains such as geology, 
climatology, sociology, epidemiology, and transportation science. This 
kind of data is distinct from other data types due to the simultaneous 
presence of spatial and temporal dimensions, which increase analysis 
complexity substantially. Purely automatic data analysis techniques, 
therefore, are insufficient to handle such complexity immaculately. Humans 
not only have inherently good senses for perceiving space and time, but 
also have creativity, flexibility, and domain expertise. Hence, an 
appropriate manner to involve these humans' traits into the automatic data 
analysis process would be tremendously helpful.

In this thesis, we introduce the basic idea of combining automatic data 
analysis techniques with interactive visualizations in the context of 
spatio-temporal data analysis and discuss the main challenges. We present 
two advanced interactive visual analysis techniques that are developed on 
the basis of intelligent mining models to demonstrate the strength of such 
a combination. In particular, we first study the problem of billboard 
location selection using massive taxi trajectory data. The problem is 
tough because it not only has vast solution search space but also involves 
multiple factors to judge the optimal. The former requires a mass of 
computing that would be too large for a human to effectively tackle alone, 
while the latter highly demands the involvement of humans as every person 
may have different criteria. To tackle the challenges, we present 
SmartAdP, a system combining a visualization-driven mining model with 
several well-designed visualization and interaction techniques, to 
facilitate creating and comparing multiple solutions in an efficient and 
human-steerable manner. Secondly, an interactive visualization system, 
TPFlow, is presented for exploring large-scale multidimensional 
spatio-temporal data. Spatio-temporal patterns at different granularity 
levels are usually hidden within different subsets of data. In TPFlow, we 
model the spatio-temporal dataset as a tensor and propose a novel 
tensor-based algorithm to support automated tensor (dataset) partitioning 
and multidimensional pattern extraction simultaneously. Built upon the 
algorithm, the TPFlow system features a novel combination of visualization 
and interaction designs to facilitate pattern discovery, comparison, and 
verification.

The effectiveness and usefulness of the above techniques are all validated 
through case studies on real-world datasets from various application 
domains and interviews of domain experts. Notice that the proposed 
techniques are not limited to the presented example application scenarios 
but can be easily adapted to other applications with similar problems as 
well.


Date:			Friday, 3 May 2019

Time:                  	10:00am - 12:00noon

Venue:                  Room CYTG003 (CYT Building)
                         (lifts 35/36)

Committee Members:	Prof. Huamin Qu (Supervisor)
 			Prof. Tin-Yau Kwok (Chairperson)
 			Prof. Cunsheng Ding
 			Dr. Qifeng Chen


**** ALL are Welcome ****