rita almeida ribeiro ca3/cts/uninova email:...
TRANSCRIPT
Fuzzy Decision Making
All you ever wanted to know about fuzzy decision making
but were afraid to ask!
Rita Almeida Ribeiro
CA3/CTS/UNINOVA
email: [email protected]
URL: www.uninova.pt/ca3/
22-10-2010
2 Out 2010
Topics
Basic Concepts of Fuzzy Set theory
Decision Making: Crisp vs Fuzzy
1. Fuzzy Optimization - Fuzzy Multiple Objective Decision
Making (FMODM)
Examples
2. Fuzzy Multiple Attribute Decision Making (FMADM)
Examples
3. Fuzzy Inference Systems
Examples
Conclusions
3 Out 2010
Decision Making Contexts (or environments)
(1) Decision making under certainty (complete information).
E.g. If employee has children he gets child support
(3) Decision making under uncertainty (least information).
E.g. If cold take the car
(2) Decision making under risk (less than complete information).
E.g. Probability of finishing the project in 10 months?
4 Out 2010
Basics on Fuzzy Set Theory [Zadeh, 65]
A fuzzy set A in X is a set of ordered pairs, A={xi, μ(xi)| xi E X}
where (xi) represents the grade of membership of xi in A.
Corresponding Discrete Fuzzy Sets:
Around3={2.1/0.1,..., 3/1,...,3.9/0.1} Cold={-10/1,...0/1,...,5/0}
Continuous Fuzzy Sets:
5 Out 2010
Basics on Fuzzy Sets Theory
Linguistic Variables
They include a set of labels (values), each corresponding to a compatibility
function (i.e. fuzzy set). The objective is to enable expressing imprecise
semantic concepts using a consistent mathematical formulation and then
perform approximate reasoning.
6 Out 2010
Basics on Fuzzy Sets Theory [Zadeh, 65]
Not Pleasant Very Hot
Negation (complement)
Modifiers (concentration): e.g. very, few, more or less, quite, many....
7 Out 2010
OPERATORS [Klir & Folger 1988]
INTERSECTION (AND)
• T-norms (e.g. min, product, drastic product...)
• Compensatory (e.g. Hammaker, Dubois&Prade, Yager...)
UNION (OR)
• T-conorms (e.g. Max, sum, drastic sum...)
• Compensatory (e.g. Dubois & Prade...)
IMPLICATION (Inference)
• Kleene, Lukasiewicz, Mandani, Sugeno....
AGGREGATION:
• Intersection, weighted average, Yager, Sugeno.....
Note: Fuzzy logic and decision making use/need these operators
8 Out 2010
FUZZY MULTICRITERIA DECISION MAKING
MCDM
MADM MODM
Objectives
Constraints
Select Optimal Decision
Maxmin Model (Bellman & Zadeh, 70) FMADM FMODM
Zimmermann, 76, 78,96
Alternatives
Attributes
Select Best Decision
FUZZY
9 Out 2010
Basics on MCDM
MODM - consists of a set of conflicting goals that usually are difficult to
achieve simultaneously. MODM deals with problems where the alternatives
are not pre-defined, so the decision-maker has to select the more
promising alternative facing the quantity of (limited) resources available.
MADM - the alternatives are pre-determined and known. The decision-
maker has to select/prioritize/rank a finite number of alternative actions.
The choice of alternatives is performed based on the attributes
classification.
“A decision process in which the goals and constraints, but not necessarily
the system under control, are fuzzy in nature” (maxmin model) [Bellman &
Zadeh 1970]
10 Out 2010
Maxmin Model (Symmetric) [Bellman and Zadeh 1970]
Consider a fuzzy set A in X is a set of ordered pairs, where (xi)
represents the grade of membership of xi in A.
Assuming a set of goals and constraints, the fuzzy decision D is,
• Hence the “optimum” is:
“No difference from goals and constraints “SYMMETRIC MODEL”
11 Out 2010
FUZZY OPTIMIZATION
Fuzzy Constraints
Fuzzy coefficients
Decision Variables Mathematical
relationships Result variables
Degrees of violation
Goal(s): Max Z
fuzzy coefficients
12 Out 2010
Fuzzy Optimization Modes of Imprecision
Fuzzy Constraint boundaries
E.g. total time of painting should be less than around 50 hours
Fuzzy goals
E.g. profits should be above $100,000.
Fuzzy coefficients in the objective function variables
E.g. price of chair2 is around $20
Fuzzy coefficients in the constraints variables
E.g. cost per hour of producing product 1 is around $10
Formally,
NON-
LINEAR
13 Out 2010
Example - Unfeasible problem
Crisp infeasible problem!
(Constraint 4)
40
10 15 20 25 30 35 40
10
15
20
25
30
35
(1)
(3)
(4)
(2)
(5)
satisfying (1) (2) (3) (5)
x 1
Fuzzy constraints Fuzzy coefficients Both
14 Out 2010
Example 2 - Unfeasible problem
10
15
20
25
30
35
40
10 15 20 25 30 35 40
(1)
(3) (4)
(2)
(5)
satisfying (1) (2) (3) and (5)
x1
x2
(1) (2)
(3)
3 fuzzy results with different fuzzifications
15 Out 2010
Optimization Resolution Algorithms
Search Techniques
SA
evolutionary
algorithms tabu search
GA Evolutionary
strategies SA/GA
Source: adapted from IEEE Computer, June 94
calculus-based guided random enumerative
16 Out 2010
FUZZY MULTICRITERIA DECISION MAKING
MCDM
MADM MODM
Objectives
Constraints
Select Optimal Decision
Maxmin Model (Bellman & Zadeh, 70) FMADM FMODM
Zimmermann, 76, 78,96
Alternatives
Attributes
Select Best Decision
FUZZY
17 Out 2010
Multiple Attribute Decision Making (FMADM)
Consequences
States of Nature
C1 C2 … Cn
Actions
A1 x
11 x
12 … x
1n
A2 x
21 x
22 … x
2n
.
.
.
.
.
.
.
.
.
…
…
…
.
.
.
Am x
m
1
xm
2
… xm
n
“Decision making is a process of choosing among alternative courses of
action for the purpose of attaining a goal or goals”
MADM - the alternatives are pre-determined and known. The decision-
maker has to select/prioritize/rank a finite number of alternative actions.
The choice of alternatives is performed based on the attributes
classification.
EXAMPLES :
Job selection (e.g. company prestige, location, salary, career prospects…)
Missile selection (e.g. velocity, range, reliability, potential... )
Evaluation of proposals (e.g.leasing of cars, project selection, statue to
choose)
18 Out 2010
Types and Challenges of MADM Problems
Main types of problems:
1. Classification
2. Selection
Main Challenges:
1. Selection of operators
2. Definition of weights
Main Methods:
1. Dominance
2. Maxmin 8. ELECTRE
3. Maximax 9. TOPSIS
4. Conjunctive/disjunctive method
5. AHP
6. Lexicographic method
7. Weigthed product or average
[Chen & Hwang, 1992]
19 Out 2010
FMADM Application: best site to land
20 Out 2010
FMADM- Example 2
Prioritization of equipment repairs under battle conditions
A Fuzzy Decision Support System for Equipment Repair under Battle Conditions. M. S. Marques, R. A. Ribeiro, A. G. Marques. Fuzzy Sets and Systems, Nr 115 (2000), pp 141-157.
21 Out 2010
Alternatives
Attributes
Select Best Decision
FMADM FMODM
Objectives
Constraints
Select Optimal Decision
Maxmin Model (Bellman & Zadeh, 70)
FIS/FKBS
Zimmermann, 76
Inputs,Outputs
Rules
Infer Best Decision
Mamdani, 74
22 Out 2010
Fuzzy Inference Systems
Fuzzy Knowledge Based System (FBKS)/Fuzzy Inference Systems (FIS)/Fuzzy Expert System (FES)
FIS
23 Out 2010
FIS: MODI
Alarm System
Input Telemetry
Nominal
Terrain Recognition
Terrain detected
No Terrain detected
Alarm
Unknown Situation
The System doesn’t
know the actual
situation.
Critical alarm
Terrain type, MPa and
Confidence-level.
Lack of information
to detect the terrain.
If alarm is not critical it
will infer about the
terrain
Alarm Reasoning
IF Set-Points are satisfied (terrain, translation, rotation) and the variables are nominal Then No-alarm.
IF Set-Points are satisfied (terrain, translation, rotation) and one or more variables have deviations Then alarm-level (critical, severe, moderate etc)
IF Set-Points are NOT satisfied (terrain, translation,rotation) Then unknown situation.
24 Out 2010
CONCLUSION
Absolute certainty is difficult to find in the real world.
Natural language is an imprecise mode of expressing concepts.
Since the majority of problems faced by decision makers are ill- structured they require specific tools and methods.
“... they have to be capable of actually making or recommending
decisions, taking as their inputs the kinds of empirical data that are
available in the real world. ... idealised models of optimising
entrepreneurs, equipped with complete certainty about the world ....
are of little use”. H. Simon
25 Out 2010
Bibliography
Bellman, R. E., and Zadeh, L. A. (1970). “Decision-Making in a Fuzzy Environment.” Management Science, Vol.17(No.4), 141-164.
Chen, S.-J. and C.-L. Hwang (1992). Fuzzy Multiple Attribute Decision Making, Springer-Verlag.
Klir, G. J. and T. A. Folger (1988). Fuzzy Sets, Uncertainty, and Information, Prentice-Hall International Editions.
Lai, Y.-J., and Hwang, C.-L. (1992). Fuzzy Mathematical Programming: Methods and Applications, Springer-Verlag.
Lai, Y.-J., and Hwang, C.-L. (1994). Fuzzy Multiple Objective Decision Making, Springer-Verlag.
Ribeiro, R. A., and Pires, F. M. (1999) “Fuzzy Linear Programming via Simulated Annealing.”Kybernetika, Vol 35, Nr 1, 57:67.
Ribeiro, Rita A. (1996). “Fuzzy Multiple Attribute Decision Making: A Review and New Preference Elicitation Techniques”, Fuzzy Sets and Systems . 78, pp 155-181.
Zadeh, L. A. (1965). "Fuzzy Sets." Information and Control 8: 338-353.
Zadeh, L. A. (1987). The Concept of a Linguistic Variable and its Application to Approximate Reasoning - I. Fuzzy Sets and Applications: Selected Papers by L. A. Zadeh. R. R. Yager, S. Ovchinnikiv, R. Tong and H. T. Nguyen. New York, John Wiley & Sons. Reprinted from: Information Sciences, 8 (1975): 219-269.
Zimmermann, H.-J. (1996). Fuzzy Set Theory and its Applications, Kluwer Academic publisher, Boston.