Perception-Based Classification (PBC) System

1
Perception-Based Classification(PBC) System

• Salvador Ledezma
• sledezma_at_uci.edu
• April 25, 2002

2
Introduction

• Concepts
• Demo of PBC
• References
• Towards and Effective Cooperation of the User
  and Computer for Classification
• Visual Data Mining with Pixel-oriented
  Visualization Techniques
• Visual Classification An Interactive Approach
  to Decision Tree Construction
• Mihael Ankerst, author or coauthor

3
Data Mining

• Exploration and Analysis, by automatic or
  semi-automatic means, of large quantities of data
  in order to discover meaningful patterns and
  rules
• Part of Knowledge Discovery in Databases (KDD)
  process

4
Classification

• Major task of Data Mining
• Assign object to one of a set of given classes
  based on object attributes

5
Classification Algorithms

• Decision Tree Classifier
• Training set set of objects whose attributes
  and class is already known
• Using training set, tree classifier determines a
  classification function represented by a decision
  tree
• Model for class attribute as a function of the
  values of other attributes
• Test set validates the classification function

6
Classification Example
7
(No Transcript)
8
Classification (cont)

• Usually algorithms are black boxes with no user
  interaction or intervention
• Reasons for user involvement in decision tree
  construction
• Use human pattern recognition capabilities
• User will have better understanding of tree
• User provides domain knowledge

9
Visual Data Mining

• Tackle data mining tasks by
• enabling human involvement
• Incorporating perceptivity of humans

10
Visual Classification

• Construction of decision trees is decomposed into
  substeps
• Enables human involvement
• Example PBC
• Data visualization based on 2 concepts
• Each attribute of training data is visualized in
  a separate part of screen
• Different class labels of training objects are
  represented by different colors

11
Pixel-Oriented Visualization Techniques

• Represent each attribute value as a single
  colored pixel
• Map the range of possible attribute values to a
  fixed color map
• Maximizes the amount of information represented
  at one time without any overlap

12
Circle Segments Technique

• Data is a circle divided into segments
• Each segment represents an attribute
• Attribute values are mapped by a single colored
  pixel and arrangement starts in the center and
  proceeds outward
• Example

13
Light high stock price Dark low stock price
Represents 50 stocks. 1 circle represents the
prices of different stocks at the same time
14
Bar Visualization

• For each attribute
• Attribute values are sorted into attribute lists
• Classes are defined by colors
• Within a bar, sorted attribute values are mapped
  to pixels, line by line
• Each attribute is placed in a different bar

15
DNA Training Data Attribute 85 and attribute 90
visually are good candidates for splitting
tree Algorithm picks 90 as the optimal split
16
PBC

• Uses pixel-oriented visualization
• Visualizes training data in order to support
  interactive decision tree construction
• Examples of use
• Automatic
• Automatic-manual (top 2 levels)
• Manual-automatic
• Manual
• Actual use lies somewhere in between this spectrum

17
(No Transcript)
18
Additional Functionality

• Propose split
• Look-ahead
• For a hypothetical split
• Expand tree
• Automatic expanding and construction

19
PBC demo