Scribd
Upload
4 views

C6- Introduction to MODM

Uploaded by

thuytien310503
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
4 views

C6- Introduction to MODM

Uploaded by

thuytien310503
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 41

MULTIPLE OBJECTIVE

DECISION MAKING
(MODM)
Ha Thi Xuan Chi, PhD

1
Contents 2

Introduction
Classification of MODM methods
Modeling
The optimal solutions and efficient
solutions.
The ideal solutions
Introduction 3

Multiple Objective Decision Making (MODM) is


not completely a minimization or maximization
problem.
MODM is a special combination among single
optimization problems, usually these objectives are
conflicting. Called “Satisficing”: satisfying and
sacrificing.
Mathematical Programming played important role
in modeling .
Classification of MODM
methods
MODELING 5

When k objectives are considered


simultaneously, the MODM problem can
be modeled as follows:

Where:
, …, )
MODELING 6

The matrix form:


maximize Z = CX
Subject to: AX  B;
X0
Where:
Z: matrix with dimension kx1;
X: matrix with dimension nx1;
B: matrix with dimension mx1;
A: matrix with dimension mxn and
C: matrix with dimension kxn.
Modeling
Example 1:

max z1 = x1 + x2
max z2 = x2–x1
Subject to:
x1  3
x2  3
x1 , x2  0

There is no optimal solution.


Modeling
Example 2:

max z1 = x2
max z2 = x2–x1
Subject to:
x1  3
x2  3
x1 , x2  0

The optimal solution is (0,3)


MODELING 9

Notes:
 It is impossible to satisfy all conflicting objectives.
 Optimality is replaced by concept “satisfying and
sacrificing” or “the best compromise solutions”.
This is based on Decision Maker preferences.

Assumptions:
 The solutions were exist.
 Objectives might not be modified and compacting
to simpler or less objectives than the original one.
Solutions
The optimal solution,
The efficient solution
The ideal solution.
Optimal solutions 11

The optimal solution: is the maximum value which achieving from


all objectives simultaneously.
Solution x* is optimal if and only if:
x* S and fj(x*)≥ fj (x)
j and x  S;
S is set of feasible solutions.
Notes:
- In general, there is no feasible optimal solution.
- If objectives are not conflicting, than one can find out optimal
solution. However, by nature, objectives are conflicting, so we
consider only efficient solutions.
Dominance
x1 dominates x2, if
Solution x1 is no worse than x2 in all
objectives
Solution x1 is strictly better than x2 in
at least one objective
x1 dominates x2 <=>x2 is dominated
by x1
Efficient solutions 13

Efficient solution (Pareto solution, frontier


solutions, non-dominant solution)
 A solution that no increase can be obtained at
any objective without simultaneously decrease at
least one of objectives.
 Solution x* is efficient if and only if there does
not exist any x  S sao cho:
fj (x)  fl (x*), for All j
fj (x) > fj (x*),for at least one value of j.
The solution is not unique.
Efficient solutions 14

Feasible solutions
Set of efficient solutions
Z1Z
=f11=
(x)f (x) Taäp caùc ñieåmhieäu quaû
aA

bb
00

Z = f (x)
Z2 =f22 (x)

Fig. 2.1 Illustration on concept of Efficient solutions


THE BEST EFFICIENT 15

SOLUTIONS
In most of the cases, we need a “best”
among efficient solutions.

Therefore, some additional criteria will


be introduced to select the “best”.

The term “best compromise solution”


will be used.
THE BEST EFFICIENT 16

SOLUTIONS
Example 3 Consider following bi-criterion problem. Find the
efficient solutions.
maximize z1 = –2x1 + x2;
maximize z2 = 2x1 + x2
Sub. to: - x1 + x2  1
x1 + x2  7
x1  5
x2  3
x1, x2  0
THE BEST EFFICIENT 17

SOLUTIONS
x 2 Z = f (x) 1 1

Z 2= f 2 (x)

(2, 3) (4, 3)
3
(5,2)

(0, 1)

0 (5,0) x1

Fig. 2.2 Efficient solutions of the problems


THE IDEAL SOLUTION 18

Each objective has individual optimum


solution. Corresponding with the
optimum of one objective, the other has
not reached optimum.
THE IDEAL SOLUTION
The payoff matrix:
 square matrix, the diagonal of matrix is
optimal values for each individual objective.

 xh* is the optimum solution for individual


objective h and this value was used to
compute values of the other objective.
THE IDEAL SOLUTION 20

Payoff Matrix:
x
z x1* x2* ... xh* ... xk*
Z1 f1(x1*) f1(x2*) ... f1(xh*) ... f1(xk*)
Z2 f2(x1*) f2(x2*) ... f2(xh*) ... f2(xk*)
    
zl fl(x1*) fl(x2*) ... fl(xh*) ... fl(xk*)

    
zk fk(x1*) fk(x2*) ... fk(xh*) ... fk(xk*)
THE IDEAL SOLUTION 21

Example 4
maximize z1 = –x1 + 3x2
maximize z2 = 2x1 + x2
maximize z3 = –2x1 + x2
Subject to:
–x1 + x2  1
x1 + x2  7
x1  5
x2  3
x1, x2  0
THE IDEAL SOLUTION 22
THE IDEAL SOLUTION 23

Individual Optimum values::


x11* = 2; x21* = 3; z1* = 7
x12* = 5; x22* = 2; z2* = 12
x13* = 0; x23* = 1; z3* = 1
The payoff matrix:
x1* x2* x3*
Z1 7 1 3
Z2 7 12 1
Z3 -1 -8 1
TECHNIQUES 24

• The method of global criterion


• Goal Programming
• Epsilon-constraint
• Denovo programming
The method of global
criterion
25

MODM problem :

Transform to the method of global criterion for the


vector maximum problem
,

The ideal solutions, (x)


The method of global
criterion
Case 1: p =1: f(x) and g(x) are linear functions
When all the objectives, i =1, 2, .., k, and all the
constraint functions, i = 1, ... , m, are linear
functions,

becomes a linear programming problem


The simplex method for L.P. can solve the problem.
The method of global
criterion
27

Case 2: p = 2: f(x) and g(x) are linear


functions
When both f(x) and g(x) are linear functions

becomes a convex programming problem; more


precisely, a quadratic programming problem.
The method of global
criterion
Case f(x) and g(x) are non-linear functions
If anyone or all of the objectives,
, j = 1, 2, ••• ,k,
and , i =1, 2, •••, m, are nonlinear functions,

is a nonlinear programming problem (for both p =1


or p =2).
Case p=:
,

Become :
Minmax d∞ =
Ex: Production scheduling of
the Hardee toy company
The Hardee toy company makes two kinds of toy dolls. Doll A
is a high quality toy and Doll B is of lower quality. The
respective profits are $0.40 and $0.30 per doll. Each Doll A
requires twice as much time as a Doll B, and if all dolls were
of type B, the company could make 500 per day. The supply
of material is sufficient for only 400 dolls per day (both A and
B combined). The problem assumes that all the dolls for type
A and type B the factory can make could be sold, and that the
best customer of the company wishes to have as many as
possible of type A doll. The manager realizes that two
objectives:
 (1) the maximization of profit, and
 (2) the maximum production of Doll A, should be considered in
scheduling the production.
Formulate the problem
Solution
x1 and x2 are the number of Doll A and
that of Doll B produced
The method of global criterion follows
three steps:
 Step 1. Obtain the ideal solution;
 Step 2. Construct a pay-off table;
 Step 3. Obtain the preferred solution.
Step 1: Obtain the ideal
solution

The solution is: x1 =100, x2= 200, f1 (x*) = 130


Step 1: Obtain the ideal
solution

The optimal solution: x1 = 250, x2 = 0, f2(x*) =250


Step 2: Construct a pay-off
table
Step 2: Construct a pay-off
table
Row 1 gives the solution vector:
x*= (100, 300) , f1(x*) = 130; f2 (100,
300) =100 is the value taken on by the
objective f2 when f1 reaches its
maximum.
Row 2 gives the solution vector:
 x* = (250, 0) , f2 (x*) = 250; f1 (250, 0)
= 100 is the value taken on by f1 when f2
reaches its maximum.
Step 3. Obtain the preferred
solution
 A preferred solution is a non-dominated
solution which is a point on the segment
of straight line BC
Step 3. Obtain the preferred
solution
Case1. P=1

The solution is given by


x1= 250,
x2= 0
f1= 100,
f2=250
which is point C
Step 3. Obtain the preferred
solution
Case 2. p = 2

The solution is :
x1= 230.7;
x2= 38.6
f1=103.9,
f2=230.7
which is point D

You might also like