Developing a diagnostic system through integration of fuzzy case-based reasoning and fuzzy ant colony system Expert Systems with Applications 28(2005)

1
Developing a diagnostic system through
integration of fuzzy case-based reasoning and
fuzzy ant colony systemExpert Systems with
Applications 28(2005)

• Author R.J. Kuo, Y.P. Kuo,
• Kai-Ying Chen
• Speaker Chih-Yao Chien

2
Outline

• Introduction
• CBR
• Fuzzy CBR
• ACS
• ASCA
• Fuzzy ant K-means algorithm
• Experiment
• QA

3
Introduction

• In order to cope with huge amount of data and
  information in the business, varieties of methods
  including artificial intelligence and statistical
  methods are developed to extract valuable
  information from the raw data.
• Case-based reasoning is one of these methods.

4
Case-based reasoning (CBR)

• CBR - Searching for similar cases from the
  historical cases for user as consulting
  references in solving needed problems.
• CBR cycle retrieve, reuse, revise, retain.

5
(No Transcript)
6
Case-based reasoning (CBR)

• Fuzzy CBR fuzzy sets theory is used for
  evaluating similarity between a new case and the
  existing cases in the case base.
• In general CBR matching mechanism, successful
  matching of a selected index is an all-or-nothing
  affair.?the interval may be too small or
  specific resulting in no matches for a given
  observed feature set.?requiring a very large
  case library to cover the input space.

7
Case-based reasoning (CBR)

• Fuzzy similarity method is proposed to improve
  the effectiveness of indexing and matching
  accuracy.

8
(No Transcript)
9
Ant colony system (ACS)
10
Ant colony system (ACS)-ASCA

• Fuzzy ant system-based clustering algorithm

11
Ant colony system (ACS)-Fuzzy AK

• Fuzzy ant K-means algorithm

12
Experiment
13
Experiment
14
Experiment

• Fuzzy sets theory indeed improves the ASCAAK
  method.
• This study has presented the capability
  advantages of using fuzzy CBR.
• Thinking style of human beings.
• Easily extract the domain knowledge experts
  know-how.
• Searching time is considerably less.

15
Experiment

• Drawbacks
• Find sufficient amount of cases.
• Most domain experts are not willing to provide
  their own know-how.
• If there are too many selections for a single
  attribute, it is possible that these selections
  are extremely close.