VisDB and Pixel Bar Charts

1
VisDB and Pixel Bar Charts

• Daniel A. Keim et al.
• ICS 280 Information Visualization
• Presented by Jeff Ridenour 4/16/02

2
Introduction

• Some problems with Standard Representations
• Limited number of attributes may be examined at
  one time
• Limited Number of Instances
• Query Specification Problem Too general or
  specific, no feedback

3
Information Visualization Techniques

• Data Type to be visualized (1D, 2D, Multi-D,
  text, hierarchies, etc.)
• Visualization Technique (standard 2D/3D,
  Geometrically Transformed Icon based, dense
  pixel, stacked)
• VisDB is dense pixel, Pixel Bar Charts mix
  standard and dense pixel.

4
VisDB

• Center is Most Relevant, spiraling out
• Color also represents relevance
• Calculating Relevance Factors, Distance Functions
• Coloration of Relevance Factors

5
Types of Display

• Basic Visualization
• 2D Arrangement
• Grouping Arrangements

6
Examples and Implementation

• Interactive Data Exploration
• Examples
• Visualizing Complex Queries
• Implementation

7
Pixel Bar Charts

• Compared to Regular Bar Charts
• Basic Idea of Pixel Bar Charts
• Space filling pixel Bar Charts
• Multi-Pixel Bar Charts

8
Formal Definition of Pixel Bar Charts

• Dividing, Ordering, and Coloring Attributes.
• Ordering Constraints ( Dense Display, No-Overlap,
  Locality, Ordering)
• Pixel Placement Algorithm (Complexity O(n logn)

9
Implementation, Application, and Evalutation

• Implemented as Java Applet, Interactive Data
  Analysis
• E-commerce example
• Comparison of Parallel Coordinate, Generalized
  Spiral (VisDB) and Pixel Bar Chart Techniques.