Visual analytics of geo-related multidimensional data

Li, M 2019, Visual analytics of geo-related multidimensional data, Doctor of Philosophy (PhD), Science, RMIT University.


Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Li.pdf Thesis application/pdf 46.56MB
Title Visual analytics of geo-related multidimensional data
Author(s) Li, M
Year 2019
Abstract In recent years, both the volume and the availability of urban data related to various social issues, such as real estate, crime and population are rapidly increasing. Analysing such urban data can help the government make evidence-based decisions leading to better-informed policies; the citizens can also benefit in many scenarios such as home-seeking. However, the analytic design process can be challenging since (i) the urban data often has multiple attributes (e.g., the distance to supermarket, the distance to work, schools zone in real estate data) that are highly related to geography; and (ii) users might have various analysis/exploration tasks that are hard to define (e.g., different home-buyers might have requirements for housing properties and many of them might not know what they want before they understand the local real estate market). In this thesis, we use visual analytics techniques to study such geo-related multidimensional urban data and answer the following research questions.

In the first research question, we propose a visual analytics framework/system for geo-related multidimensional data. Since visual analytics and visualization designs are highly domain-specific, we use the real estate domain as an example to study the problem. Specifically, we first propose a problem abstraction to satisfy the requirements from users (e.g., home buyers, investors). Second, we collect, integrate and clean the last ten year's real estate sold records in Australia as well as their location-related education, facility and transportation profiles, to generate a real multi-dimensional data repository. Third, we propose an interactive visual analytic procedure to help less informed users gradually learn about the local real estate market, upon which users exploit this learned knowledge to specify their personalized requirements in property seeking. Fourth, we propose a series of designs to visualize properties/suburbs in different dimensions and different granularity. Finally, we implement a system prototype for public access (http://115.146.89.158), and present case studies based on real-world datasets and real scenario to demonstrate the usefulness and effectiveness of our system.

Our second research question extends the first one and studies the scalability problem to support cluster-based visualization for large-scale geo-related multidimensional data. Particularly, we first propose a design space for cluster-based geographic visualization. To calculate the geographic boundary of each cluster, we propose a concave hull algorithm which can avoid complex shapes, large empty area inside the boundary and overlaps among different clusters. Supported by the concave hull algorithm, we design a cluster-based data structure named ConcaveCubes to efficiently support interactive response to users' visual exploration on large-scale geo-related multidimensional data. Finally, we build a demo system (http://115.146.89.158/ConcaveCubes) to demonstrate the cluster-based geographic visualization, and present extensive experiments using real-world datasets and compare ConcaveCubes with state-of-the-art cube-based structures to verify the efficiency and effectiveness of ConcaveCubes.

The last research question studies the problem related to visual analytics of urban areas of interest (AOIs), where we visualize geographic points that satisfy the user query as a limited number of regions (AOIs) instead of a large number of individual points (POIs). After proposing a design space for AOI visualization, we design a parameter-free footprint method named AOI-shapes to effectively capture the region of an AOI based on POIs that satisfy the user query and those that do not. We also propose two incremental methods which generate the AOI-shapes by reusing previous calculations as per users' update of their AOI query. Finally, we implement an online demo (http://www.aoishapes.com) and conduct extensive experiments to demonstrate the efficiency and effectiveness of the proposed AOI-shapes.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Science
Subjects Global Information Systems
Database Management
Keyword(s) Geographic Visualization
Multidimensional Visualization
Visual Analytics
Data Reduction
Footprint Algorithm
Visualization Scalability
Versions
Version Filter Type
Access Statistics: 47 Abstract Views, 40 File Downloads  -  Detailed Statistics
Created: Wed, 05 Jun 2019, 13:32:26 EST by Adam Rivett
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us