Slide index
Rough Sets in
KDD Tutorial Notes
About the Speakers
About the Speakers
(2)
Contents
Introduction
Introduction (2)
Introduction (3)
Basic Concepts of
Rough Sets
Information
Systems/Tables
Decision
Systems/Tables
Issues in the
Decision Table
Indiscernibility
Indiscernibility (2)
An Example of
Indiscernibility
Observations
Set Approximation
Set Approximation
(2)
An Example of Set
Approximation
An Example of
Set Approximation (2)
slide PPT
slide PPT
slide PPT
slide PPT
Properties of
Approximations
Properties of
Approximations (2)
Four Basic Classes
of Rough Sets
Accuracy of
Approximation
Issues in the
Decision Table
Reducts
Dispensable &
Indispensable Attributes
Independent
Reduct & Core
An Example of
Reducts & Core
Discernibility
Matrix (relative to positive region)
Discernibility
Matrix (relative to positive region) (2)
Discernibility
Function (relative to objects)
Examples of
Discernibility Matrix
Examples of
Discernibility Matrix (2)
Rough Membership
Rough Membership (2)
Dependency of
Attributes
Dependency of
Attributes (2)
Dependency of
Attributes (3)
A Rough Set Based
KDD Process
What Are Issues of
Real World ?
slide PPT
slide PPT
slide PPT
slide PPT
slide PPT
A Rough Set Based
KDD Process
Observations
Discretization based
on RSBR
Discretization Based
on RSBR (2)
A Geometrical
Representation of Data
A Geometrical
Representation of Data and Cuts
Discretization Based
on RSBR (3)
Discretization Based
on RSBR (4)
A Discretization
Process
The Set of Cuts on
Attribute a
A Discretization
Process (2)
A Sample Defined in
Step 2
The Discernibility
Formula
The Discernibility
Formulae for All Different Pairs
The Discernibility
Formulae for All Different Pairs (2)
A Discretization
Process (3)
The Discernibility
Formula in CNF Form
The Discernibility
Formula in DNF Form
The Minimal Set Cuts
for the Sample DB
A Result
A Rough Set Based
KDD Process
Observations
The Goal of
Attribute Selection
Attribute Selection
The Filter Approach
The Wrapper Approach
Basic Ideas:
Attribute Selection using RSH
Why Heuristics ?
The Rule Selection
Criteria in GDT-RS
Attribute Evaluation
Criteria
Main Features of RSH
An Example of
Attribute Selection
slide PPT
slide PPT
slide PPT
slide PPT
Searching for CORE
(5)
R={b}
Attribute Evaluation
Criteria
Selecting Attribute
from {a,c,d}
Selecting Attribute
from {a,c,d} (2)
Selecting Attribute
from {a,c,d} (3)
Selecting Attribute
from {a,c,d} (4)
A Heuristic
Algorithm for Attribute Selection
A Heuristic
Algorithm for Attribute Selection (2)
A Heuristic
Algorithm for Attribute Selection (3)
Experimental Results
A Rough Set Based
KDD Process
Main Features of
GDT-RS
A Sample DB
A Sample GDT
Explanation for GDT
Probabilistic
Relationship Between PIs and PGs
Unseen Instances
Rule Representation
Rule Strength (1)
Rule Strength (2)
Rule Strength (3)
Rule Discovery by
GDT-RS
Regarding the
Instances (Noise Rate = 0)
Generating
Discernibility Vector for u2
Obtaining Reducts
for u2
Generating Rules
from u2
Generating Rules
from u2 (2)
Generating
Discernibility Vector for u4
Obtaining Reducts
for u4
Generating Rules
from u4
Generating Rules
from u4 (2)
Generating Rules
from All Instances
The Rule Selection
Criteria in GDT-RS
Generalization
Belonging to Class y
Generalization
Belonging to Class n
Results from the
Sample DB (Noise Rate = 0)
Results from the
Sample DB (2) (Noise Rate > 0)
Regarding
Instances (Noise Rate > 0)
Rules Obtained from
All Instacnes
Example of Using BK
Changing Strength of
Generalization by BK
Algorithm
1 Optimal Set of Rules
Algorithm
1 Optimal Set of Rules (2)
Algorithm
1 Optimal Set of Rules (3)
The Issue of
Algorithm 1
Algorithm 2
Sub-Optimal Solution
Algorithm
2 Sub-Optimal Solution (2)
Algorithm
2 Sub-Optimal Solution (3)
Algorithm
2 Sub-Optimal Solution (4)
Time Complexity of
Alg.1&2
Experiments
Experiment
1 (meningitis data)
Experiment
1 (meningitis data) (2)
Experiment
1 (meningitis data) (3)
Using Background
Knowledge (meningitis data)
Using Background
Knowledge (meningitis data) (2)
Explanation of BK
Using Background
Knowledge (meningitis data) (3)
Using Background
Knowledge (meningitis data) (4)
Experiment
2 (bacterial examination data)
Attribute
Selection (bacterial examination data)
Some Results
(bacterial examination data)
Experiment
3 (gastric cancer data)
Result of Attribute
Selection (gastric cancer data)
Result of Attribute
Selection (2) (gastric cancer data)
Experiment
4 (slope-collapse data)
Result of Attribute
Selection (slope-collapse data)
The Discovered Rules
(slope-collapse data)
Other Methods for
Attribute Selection (download from
http://www.iscs/nus.edu.sg/liuh/)
slide PPT
slide PPT
Result of WSFG
Result of WSFG
(2) (class: direct death)
Result of WSBG
Result of WSBG
(2) (class: direct death)
Result of
LVI (gastric cancer data)
Some Rules
Related to Direct Death
GDT-RS vs.
Discriminant Analysis
GDT-RS vs. ID3
(C4.5)
Rough Sets in ILP
and GrC -- An Advanced Topic --
Advantages of ILP
(Compared with Attribute-Value Learning)
Weak Points of
ILP (Compared with Attribute-Value Learning)
Goal
Normal Problem
Setting for ILP
Normal Problem
Setting for ILP (2)
Normal Problem
Setting for ILP (3)
Issues
Imperfect Data in
ILP
Imperfect Data in
ILP (2)
Imperfect Data in
ILP (3)
Observations
Observations (2)
Solution
Why GrC? A
Practical Point of View
Solution (2)
Rough Sets
Rough Sets (2)
Rough Sets (3)
An Illustrating
Example
An Illustrating
Example (2)
Rough Problem
Setting for Insufficient BK
Rough Problem
Setting for Insufficient BK (2)
Rough Problem
Setting for Insufficient BK (3)
Rough Problem
Setting for Insufficient BK (4)
Example Revisited
Example Revisited
(2)
Example Revisited
(3)
Rough Problem
Setting for Indiscernible Examples
Rough Problem
Setting for Indiscernible Examples (2)
Rough Sets (GrC) for
Other Imperfect Data in ILP
Future Work on RS
(GrC) in ILP
Summary
Advanced
Topics (to deal with real world problems)
Advanced Topics
(2) (to deal with real world problems)
References and
Further Readings
References and
Further Readings
References and
Further Readings
References and
Further Readings
References and
Further Readings
References and
Further Readings
References and
Further Readings
Related Conference
and Web Pages
slide PPT
|
Info:
The presentation is quite large. It contains 210 slides.
The HTML version is optimised for 800x600 resolution,
full color. The PowerPoint version of slides can be
downloaded by clicking on the link below.
The size of .ppt file is approximately 1,2 MB.
Download
Please take care of copyrights while passing this
material to third-party.
|