You are on page 1of 27

Case-Based Reasoning

Faced this problem before?


Oops the car stopped. What could have gone
wrong?
Aah Last time it happened, there was no petrol. Is
there petrol?
Yes. Oh but wait I remember the tyre was
punctured.
This is the normal thought process of a human
when faced with a problem which is similar to a
problem he/she had faced before.

What do we do?
Reuse the solution experience when faced
with a similar problem.
What if we extend this to problem solving
when using machine?
We know are talking of CBR.

Whats CBR?
To solve a new problem by remembering a previous
similar situation and by reusing information and
knowledge of that situation
E.g.:
A doctor remember his previous case to diagnose and treat his
patient
A financial consultant working on a difficult credit decision
task, uses a reminding to a previous case, which involved a
company in similar trouble as the current one, to recommend
that the loan application should be refused.

History
Roots of CBR is found in the works of Roger
Shank on dynamic memory.
Other trails into the CBR field has come from
Analogical reasoning
Problem solving and experimental learning within
philosophy and psychology

The first CBR system, CYRUS, by Janet Kolodner


at Yale university.

Some notable CBR systems

CYRUS
PROTOS
HYPO
MEDIATOR
CLAVIER
CASEY
JULIA

Popularity of CBR
It is on the increase
Work on in over 35 universities
We have over 17 commercially available too
ls
A number of papers and conferences.

Fundamentals of CBR
Identify the current problem situation
Find a past case similar to the new one - RETRIEVE
Use that case to suggest a solution to the current
problem - REUSE
Evaluate the proposed solution - REVISE
Update the system by learning from this experience RETAIN

CBR Cycle

An Example

Technical Diagnosis of Car Faults

Existing Cases

Case Representation
Case-based reasoner depends heavily on the
structure and content of its collection of
cases.
Since a problem is solved by recalling a
previous experience suitable for solving the
current problem, the case search and
matching process has to be effective and
reasonably time efficient.

Case Representation
Flat feature-value list
Object Oriented representation
Graph representation
The choice of representation is
Dependent on requirements of domain and task
Structure of already available case data

Problem to be solved

Figure Credit: R. Bergmann, University of Kaiserslautern

Case Retrieval
Retrieval starts with a (partial) problem
description, and ends when a best matching
previous case has been found.

Similarity Measurement
Purpose: To select the most relevant case
Basic Assumption: Similar problems have similar
solutions
Similarity value between 0 and 1 are assigned for
feature value pairs
E.g.: Feature: Problem
Front Light does not work

.8 work
Break Light does not

Front Light does not work


.4
Engine doesnt start

Similarity Measurement
Feature: Battery Voltage
12.6

.9

13.6

12.6

.1

6.7

Different features have different importance


Two kinds of Similarity Measures
Local Similarity similarity on feature level
Global Similarity - similarity on case or object
level

Similarity Computation for case 1

Figure Credit: R. Bergmann, University of Kaiserslautern

Similarity Computation for case 2

Figure Credit: R. Bergmann, University of Kaiserslautern

Solution to the new case


Diagnosis: Break Light fuse defect
Repair : Replace break light fuse.

Retrieval
Different approaches depending on
Case representation
Case base size

Organization of case base


Linear List small case base
Index structures for large case bases
Kd-trees
Retrieval Nets
Discrimination Nets

How to store case


Database
Main Memory

Reuse
Copy
Manual/Interactive solution adaptation by
user
Automatic solution adaptation
Transformation analogy
Derivational Analogy
Compositional adaptation

Revise
No Revision
Verification of solution by simulation
Verification/evaluation in real world situation
Criteria
Correctness of solution
Quality of the solution
Other, e.g. user preference

Retain
Learn a new case
Solve a similar problem better
Issues
Size of case base
Deleting cases not used

Where does a CBR system store


its Knowledge?

Vocabulary
Case base
Similarity assessment
Solution adaptation

Advantage of CBR
Reduces knowledge acquisition effort
Requires less maintenance effort
Improve problem solving performance by
reuse
Makes use of existing data
Improve over time and adapt to changes in
the environment
Higher user acceptance

References
Case-Based Reasoning Janet Kolodner
http://www.cbr-web.org/CBR-Web/cbrintro/index.html Dr Ralph Bergmann.
Case-Based Reasoning: Foundational Issues,
Methodological Variations, and System Approaches
Agnar Aamodt, University of Trondheim; Enric Plaza,
Insitut dInvestigacio en Intel-ligencia.
Case Based Reasoning M Sasikumar, NCST.

You might also like