VISUAL ANALYTICS OF BI-DIRECTIONAL AND CLUSTER MOVEMENTS

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


PhD Thesis Defence

Title: "VISUAL ANALYTICS OF BI-DIRECTIONAL AND CLUSTER MOVEMENTS"

By

Mr. Yixian ZHENG


Abstract

Movement is a fundamental phenomenon exists ubiquitously in our daily life. 
Many crucial scientific, societal and commercial decisions are made depending 
on proper knowledge and correct understanding of movement patterns of people, 
animals or objects. However, analyzing and exploring movement is not an easy 
task due to its intrinsic multi-variate natures, hidden correlations among 
properties and complex analytical tasks in real world applications. Therefore, 
analysts seek the help of visualization to integrate humans in the data 
exploration process, applying their perceptual abilities to target datasets and 
leveraging their domain knowledge to guide the exploration for effective 
understanding, reasoning and decision making. Movement visualization has been 
widely studied in recent years concerning about novel and effective visual 
designs and user interactions for analysts to gain insight into real world 
datasets. In this thesis, we follow this line of research and propose three 
visualization techniques or visual analysis approaches to facilitate in-depth 
analyses of bi-directional and cluster movements.

In the first work, we propose a visual analytics system to investigate 
bi-directional movement behaviors. More specifically, we first design a 
movement model with modular DoI specification characterizing bi-directional 
movement. Then several novel visualization designs, including ODpair Flow View 
and Isotime Storyline View are proposed with intuitive user interactions to 
allow users to interactively explore and analyze both micro and macro patterns 
of bi-directional movement behaviors.

In the second work, we develop TelcoFlow, a visual analytics system to study 
collective behaviors based on the large scale of telco data. Advanced 
quantitative analyses including state-based probabilistic behavior model and 
biclustering are utilized to quantify and detect collective behaviors. 
Meanwhile, a set of intuitive visualization techniques with new designs is 
integrated to present the detected patterns for an in-depth analysis.

In the third work, we present a novel animation planning technique, namely 
Focus+Context Grouping, that can allow users to track movement of focal targets 
while not neglecting the context (i.e. the overall moving trend). It can be 
integrated with static visualizations to provide a more straightforward 
presentation of movement and improve users