Have a Question?
Ask the Graduate
College at our new
Doctoral Dissertation Announcement
Candidate: Mustafa Sanver
Doctor of Philosophy
Department: Computer Science
Title: Interactive Visual Exploration of Large Relational Data
Dr. Li Yang, Chair
Dr. Elise de Doncker
Dr. Karlis Kaugars
Dr. Bernard Han
Date: Thursday, March 20, 2008 3:00 p.m. – 5:00 p.m.
Parkview Campus, Room D-208
As data become ubiquitous and immense in many activities from business decision making to scientific analysis, effective analysis of such data becomes an important research area. Data analysis requires active involvement of human beings and visualization is a powerful tool to capture profound insights from data. However, interactive visual exploration of massive data poses fundamental technical challenges to both data visualization and database management systems. Billions of data records clutter the screen and existing database management systems are inadequate for overview-and-drill-down data access for interactive data exploration.
We have proposed a density-based methodology to address the above challenge. We present multiresolution data aggregation as an intermediate representation of data between visualization tools and databases. Data aggregated at multiple resolutions are stored in internal nodes of a partition-based high dimensional tree index while the individual records are stored in leaf nodes. Such a piggyback ride of aggregated data efficiently supports resolution-based data access patterns of large relational data. Thus, multiresolution data aggregation conveys density-based data input to visualizations. Therefore, many existing visualization techniques which are inefficient in visualizing large data can accept this new data input and participate in analysis of large data.
To demonstrate the feasibility and effectiveness of the new representation, a client-server visualization tool is developed. The server, KDBMS, is a kdB-tree variant to organize data in multiple resolutions and the client, mVis, is a dynamic, flexible, and extendible visualization framework. We have extended 3D footprint splatting with grand tour, density-based parallel coordinates, and density-based scatterplot matrices to render data aggregations. The framework provides a data-independent and plug-in-based environment for quick development and integration of new visualization techniques, in addition to the linking mechanism to interconnect its components.