Goal Programming

© 2003 Thomson/South-Western 1
Slide
 Goal Programming

 Goal programming may be used to solve linear

programs with multiple objectives, with each
objective viewed as a "goal".
 In goal programming, di+ and di- , deviation variables,
are the amounts a targeted goal i is overachieved or
underachieved, respectively.
 The goals themselves are added to the constraint set
with di+ and di- acting as the surplus and slack
variables.
 One approach to goal programming is to satisfy goals
in a priority sequence. Second-priority goals are
pursued without reducing the first-priority goals, etc.

© 2003 Thomson/South-Western 2
Slide
 Goal Programming

 For each priority level, the objective function is to

minimize the (weighted) sum of the goal deviations.
 Previous "optimal" achievements of goals are added to
the constraint set so that they are not degraded while
trying to achieve lesser priority goals.

© 2003 Thomson/South-Western 3
Slide
 Goal Programming Approach

Step 1: Decide the priority level of each goal.

Step 2: Decide the weight on each goal.
If a priority level has more than one goal, for
each goal i decide the weight, wi , to be placed
on the deviation(s), di+ and/or di-, from the goal.
Step 3: Set up the initial linear program.
Min w1d1+ + w2d2-
s.t. Functional Constraints,
and Goal Constraints
Step 4: Solve the current linear program.
If there is a lower priority level, go to step 5.
Otherwise, a final solution has been reached.

© 2003 Thomson/South-Western 4
Slide
 Goal Programming Approach

Step 5: Set up the new linear program.

Consider the next-lower priority level goals
and formulate a new objective function based on
these goals. Add a constraint requiring the
achievement of the next-higher priority level goals to
be maintained. The new linear program might
be:
Min w3d3+ + w4d4-
s.t. Functional Constraints,
Goal Constraints, and
w1d1+ + w2d2- = k
Go to step 4. (Repeat steps 4 and 5 until all
priority levels have been examined.)
© 2003 Thomson/South-Western 5
Slide
 Example: Innex Corporation

Innex Corporation manufactures two

products, A and B. Product A requires five hours on
machine 1 and two hours on machine 2; product B
requires two hours on machine 1 and four hours on
machine 2. The weekly capacities of machine 1 and 2
are fifty hours and forty-eight hours, respectively.

© 2003 Thomson/South-Western 6
Slide
 Example: Innex Corporation

The company has three goals which are given below:

Priority 1: Minimize underachievement of a total

production of ten units per week
(Goal 1) Priority 2: Minimize underachievement of
producing 8 units of product A weekly.
(Goal 2)
Priority 3: Minimize underachievement of
producing thirteen units of
product B weekly. (Goal 3)

© 2003 Thomson/South-Western 7
Slide
 Example: Innex Corporation

 Variables
x1 = number of units of A produced weekly
x2 = number of units of B produced weekly
di- = amount the right hand side of goal i is deficient
di+ = amount the right hand side of goal i is exceeded
 Constraints
5x1 + 2x2 < 50
2x1 + 4x2 < 48

© 2003 Thomson/South-Western 8
Slide
 Example: Innex Corporation

 Goals
(1) 10 total production weekly:
x1 + x2 + d1- - d1+ = 10
(2) 8 units product A weekly:
x1 + d2- - d2+ = 8
(3) 13 units product B weekly:
x2 + d3- - d3+ = 13
Non-negativity:
x1, x2, di-, di+ > 0 for all i

© 2003 Thomson/South-Western 9
Slide
 Example: Innex Corporation

 Formulation Summary
Min P1(d1-) + P2(d2-) + P3(d3-)
s.t. 5x1 +2x2 < 50
2x1 +4x2 < 48
x1 + x2 +d1- -d1+ = 10
x1 +d2- -d2+ = 8
x2 +d3- -d3+ = 13
x1, x2, d1-, d1+, d2-, d2+, d3-, d3+ > 0

© 2003 Thomson/South-Western 10
Slide
 Example: Innex Corporation

 Constraints and Goal Graphed

x2

25 5x1 + 2x2 < 50

x1 = 8
x2 = 13

13
x1 + x2 = 10

(8,5)
2x1 + 4x2 < 48

x1
8 10 25

© 2003 Thomson/South-Western 11
Slide
 Example: Conceptual Products

Conceptual Products is a computer company

that produces the CP400 and the CP500 computers.
The computers use different mother boards
produced in abundant supply by the company, but
use the same cases and disk drives. The CP400
models use two floppy disk drives and no zip disk
drives whereas the CP500 models use one floppy
disk drive and one zip disk drive.

© 2003 Thomson/South-Western 12
Slide
 Example: Conceptual Products

The disk drives and cases are bought from

vendors. There are 1000 floppy disk drives, 500 zip
disk drives, and 600 cases available to Conceptual
Products on a weekly basis. It takes one hour to
manufacture a CP400 and its profit is $200 and it takes
one and one-half hours to manufacture a CP500 and its
profit is $500.

© 2003 Thomson/South-Western 13
Slide
 Example: Conceptual Products

The company has three goals which are given below:

Priority 1: Meet a state contract of 200 CP400

machines weekly. (Goal 1)
Priority 2: Make at least 500 total computers weekly.
(Goal 2)
Priority 3: Make at least $250,000 weekly. (Goal 3)

© 2003 Thomson/South-Western 14
Slide
 Example: Conceptual Products

 Variables
x1 = number of CP400 computers produced weekly
x2 = number of CP500 computers produced weekly
di- = amount the right hand side of goal i is deficient
di+ = amount the right hand side of goal i is exceeded
 Functional Constraints
Availability of floppy disk drives: 2x1 + x2 < 1000
Availability of zip disk drives: x2 < 500
Availability of cases: x1 + x2 < 600

© 2003 Thomson/South-Western 15
Slide
 Example: Conceptual Products

 Goals
(1) 200 CP400 computers weekly:
x1 + d1- - d1+ = 200
(2) 500 total computers weekly:
x1 + x2 + d2- - d2+ = 500
(3) $250(in thousands) profit:
.2x1 + .5x2 + d3- - d3+ = 250
Non-negativity:
x1, x2, di-, di+ > 0 for all i

© 2003 Thomson/South-Western 16
Slide
 Example: Conceptual Products

 Objective Functions

Priority 1: Minimize the amount the state contract is

not met: Min d1-
Priority 2: Minimize the number under 500
computers produced weekly:
Min d2-
Priority 3: Minimize the amount under $250,000
earned weekly: Min d3-

© 2003 Thomson/South-Western 17
Slide
 Example: Conceptual Products

 Formulation Summary
Min P1(d1-) + P2(d2-) + P3(d3-) + P4(d4+)
s.t. 2x1 +x2 <
1000
+x2 < 500
x1 +x2 <
600
x1 +d1- -d1+ = 200
x1 +x2 +d2- -d2+ = 500
.2x1+ .5x2 +d3- -d3+ = 250
x1, x2, d1-, d1+, d2-, d2+, d3-, d3+, d4-, d4+ > 0
© 2003 Thomson/South-Western 18
Slide
 Example: Conceptual Products

 Graphical Solution, Iteration 1

To solve graphically, first graph the functional
constraints. Then graph the first goal: x1 = 200. Note on
the next slide that there is a set of points that exceed
x1 = 200 (where d1- = 0).

© 2003 Thomson/South-Western 19
Slide
 Example: Conceptual Products

 Functional Constraints and Goal 1 Graphed

x2

1000 2x1 + x2 < 1000

800 Goal 1: x1 > 200

x2 < 500
600 x1 + x2 < 600

400 Points Satisfying

Goal 1
200

x1
200 400 600 800 1000 1200

© 2003 Thomson/South-Western 20
Slide
 Example: Conceptual Products

 Graphical Solution, Iteration 2

Now add Goal 1 as x1 > 200 and graph Goal 2:
x1 + x2 = 500. Note on the next slide that there is still a
set of points satisfying the first goal that also satisfies
this second goal (where d2- = 0).

© 2003 Thomson/South-Western 21
Slide
 Example: Conceptual Products

 Goal 1 (Constraint) and Goal 2 Graphed

x2

1000 2x1 + x2 < 1000

800 Goal 1: x1 > 200

x2 < 500
600 x1 + x2 < 600

400 Points Satisfying Both

Goals 1 and 2
200
Goal 2: x1 + x2 > 500

x1
200 400 600 800 1000 1200

© 2003 Thomson/South-Western 22
Slide
 Example: Conceptual Products

 Graphical Solution, Iteration 3

Now add Goal 2 as x1 + x2 > 500 and Goal 3:
.2x1 + .5x2 = 250. Note on the next slide that no points
satisfy the previous functional constraints and goals
and satisfy this constraint.
Thus, to Min d3-, this minimum value is achieved
when we Max .2x1 + .5x2. Note that this occurs at x1 =
200 and x2 = 400, so that .2x1 + .5x2 = 240 or d3- = 10.

© 2003 Thomson/South-Western 23
Slide
 Example: Conceptual Products

 Goal 2 (Constraint) and Goal 3 Graphed

x2

1000 2x1 + x2 < 1000

800 Goal 1: x1 > 200

x2 < 500
600 x1 + x2 < 600 (200,400)

Points Satisfying Both

400
Goals 1 and 2

200 Goal 2: x1 + x2 > 500

Goal 3: .2x1 + .5x2 = 250
x1
200 400 600 800 1000 1200

© 2003 Thomson/South-Western 24
Slide
 A Scoring Model for Job Selection

 A graduating college student with a double major in

Finance and Accounting has received the following
three job offers:
•financial analyst for an investment firm in Chicago
•accountant for a manufacturing firm in Denver
•auditor for a CPA firm in Houston

© 2003 Thomson/South-Western 25
Slide
 A Scoring Model for Job Selection

 The student made the following comments:

•“The financial analyst position provides the best
opportunity for my long-run career advancement.”
•“I would prefer living in Denver rather than in
Chicago or Houston.”
•“I like the management style and philosophy at the
Houston CPA firm the best.”
 Clearly, this is a multicriteria decision problem.

© 2003 Thomson/South-Western 26
Slide
 A Scoring Model for Job Selection

 Considering only the long-run career advancement

criterion:
•The financial analyst position in Chicago is the best
decision alternative.
 Considering only the location criterion:
•The accountant position in Denver is the best
decision alternative.
 Considering only the style criterion:
•The auditor position in Houston is the best
alternative.

© 2003 Thomson/South-Western 27
Slide
 A Scoring Model for Job Selection

 Steps Required to Develop a Scoring Model

Step 1: List the decision-making criteria.
Step 2: Assign a weight to each criterion.
Step 3: Rate how well each decision alternative
satisfies each criterion.
Step 4: Compute the score for each decision
alternative.
Step 5: Order the decision alternatives from highest
score to lowest score. The alternative
with the highest score is the recommended
alternative.

© 2003 Thomson/South-Western 28
Slide
 A Scoring Model for Job Selection

 Mathematical Model

Sj = wi rij
i

where:
rij = rating for criterion i and decision alternative j
Sj = score for decision alternative j

© 2003 Thomson/South-Western 29
Slide
 A Scoring Model for Job Selection

 Step 1: List the criteria (important factors).

•Career advancement
•Location
•Management
•Salary
•Prestige
•Job Security
•Enjoyable work

© 2003 Thomson/South-Western 30
Slide
 A Scoring Model for Job Selection

 Five-Point Scale Chosen for Step 2

Importance Weight
Very unimportant 1
Somewhat unimportant 2
Average importance 3
Somewhat important 4
Very important 5

© 2003 Thomson/South-Western 31
Slide
 A Scoring Model for Job Selection

 Step 2: Assign a weight to each criterion.

Criterion Importance Weight

Career advancement Very important 5
Location Average importance 3
Management Somewhat important 4
Salary Average importance 3
Prestige Somewhat unimportant 2
Job security Somewhat important 4
Enjoyable work Very important 5

© 2003 Thomson/South-Western 32
Slide
 A Scoring Model for Job Selection

 Nine-Point Scale Chosen for Step 3

Level of Satisfaction Rating

Extremely low 1
Very low 2
Low 3
Slightly low 4
Average 5
Slightly high 6
High 7
Very high 8
Extremely high 9

© 2003 Thomson/South-Western 33
Slide
 A Scoring Model for Job Selection

 Step 3: Rate how well each decision alternative satisfies

each criterion.

Decision Alternative
Analyst Accountant Auditor
Criterion Chicago Denver Houston
Career advancement 8 6 4
Location 3 8 7
Management 5 6 9
Salary 6 7 5
Prestige 7 5 4
Job security 4 7 6
Enjoyable work 8 6 5

© 2003 Thomson/South-Western 34
Slide
 A Scoring Model for Job Selection

 Step 4: Compute the score for each decision alternative.

Decision Alternative 1 - Analyst in Chicago

Criterion Weight (wi ) Rating (ri1) wiri1

Career advancement 5 x 8 = 40
Location 3 3 9
Management 4 5 20
Salary 3 6 18
Prestige 2 7 14
Job security 4 4 16
Enjoyable work 5 8 40
Score 157

© 2003 Thomson/South-Western 35
Slide
 A Scoring Model for Job Selection

 Step 4: Compute the score for each decision alternative.

Sj = wi rij
i

S1 = 5(8)+3(3)+4(5)+3(6)+2(7)+4(4)+5(8) = 157

S2 = 5(6)+3(8)+4(6)+3(7)+2(5)+4(7)+5(6) = 167

S3 = 5(4)+3(7)+4(9)+3(5)+2(4)+4(6)+5(5) = 149

© 2003 Thomson/South-Western 36
Slide
 A Scoring Model for Job Selection

 Step 4: Compute the score for each decision alternative.

Decision Alternative
Analyst Accountant Auditor
Criterion Chicago Denver Houston
Career advancement 40 30 20
Location 9 24 21
Management 20 24 36
Salary 18 21 15
Prestige 14 10 8
Job security 16 28 24
Enjoyable work 40 30 25
Score 157 167 149

© 2003 Thomson/South-Western 37
Slide
 A Scoring Model for Job Selection

 Step 5: Order the decision alternatives from highest

score to lowest score. The alternative with the
highest score is the recommended alternative.

•The accountant position in Denver has the highest

score and is the recommended decision alternative.
•Note that the analyst position in Chicago ranks first
in 4 of 7 criteria compared to only 2 of 7 for the
accountant position in Denver.
•But when the weights of the criteria are considered,
the Denver position is superior to the Chicago job.

© 2003 Thomson/South-Western 38
Slide
 End of Goal Programming

© 2003 Thomson/South-Western 39
Slide
 Markov Process

© 2003 Thomson/South-Western 40
Slide
 Markov Processes

 Markov process models are useful in studying the

evolution of systems over repeated trials or
sequential time periods or stages.
 They have been used to describe the probability that:
•a machine that is functioning in one period will
continue to function or break down in the next
period.
•A consumer purchasing brand A in one period
will purchase brand B in the next period.

© 2003 Thomson/South-Western 41
Slide
 Transition Probabilities

 Transition probabilities govern the manner in which

the state of the system changes from one stage to the
next. These are often represented in a transition
matrix.

© 2003 Thomson/South-Western 42
Slide
 Transition Probabilities

 A system has a finite Markov chain with stationary

transition probabilities if:
•there are a finite number of states,
•the transition probabilities remain constant from
stage to stage, and
•the probability of the process being in a particular
state at stage n+1 is completely determined by the
state of the process at stage n (and not the state at
stage n-1). This is referred to as the memory-less
property.

© 2003 Thomson/South-Western 43
Slide
 Steady-State Probabilities

 The state probabilities at any stage of the process can

be recursively calculated by multiplying the initial
state probabilities by the state of the process at stage
n.
 The probability of the system being in a particular
state after a large number of stages is called a steady-
state probability.

© 2003 Thomson/South-Western 44
Slide
 Steady-State Probabilities

 Steady state probabilities can be found by solving the

system of equations P =  together with the condition
for probabilities that i = 1.
•Matrix P is the transition probability matrix
•Vector  is the vector of steady state probabilities.

© 2003 Thomson/South-Western 45
Slide
 Absorbing States

 An absorbing state is one in which the probability that

the process remains in that state once it enters the state
is 1.
 If there is more than one absorbing state, then a steady-
state condition independent of initial state conditions
does not exist.

© 2003 Thomson/South-Western 46
Slide
 Transition Matrix with Submatrices

 If a Markov chain has both absorbing and

nonabsorbing states, the states may be rearranged so
that the transition matrix can be written as the
following composition of four submatrices: I, 0, R,
and Q:

I 0

R Q

© 2003 Thomson/South-Western 47
Slide
 Transition Matrix with Submatrices

I = an identity matrix indicating one always remains

in an absorbing state once it is reached
0 = a zero matrix representing 0 probability of
transitioning from the absorbing states to the
nonabsorbing states
R = the transition probabilities from the
nonabsorbing states to the absorbing states
Q = the transition probabilities between the
nonabsorbing states

© 2003 Thomson/South-Western 48
Slide
 Fundamental Matrix

 The fundamental matrix, N, is the inverse of the

difference between the identity matrix and the Q
matrix:
N = (I - Q )-1

© 2003 Thomson/South-Western 49
Slide
 NR Matrix

 The NR matrix is the product of the fundamental (N)

matrix and the R matrix.
 It gives the probabilities of eventually moving from
each nonabsorbing state to each absorbing state.
 Multiplying any vector of initial nonabsorbing state
probabilities by NR gives the vector of probabilities for
the process eventually reaching each of the absorbing
states. Such computations enable economic analyses of
systems and policies.

© 2003 Thomson/South-Western 50
Slide
 Example: North’s Hardware

Henry, a persistent salesman, calls North's

Hardware Store once a week hoping to speak with
the store's buying agent, Shirley. If Shirley does not
accept Henry's call this week, the probability she will
do the same next week is .35. On the other hand, if
she accepts Henry's call this week, the probability she
will not do so next week is .20.

© 2003 Thomson/South-Western 51
Slide
 Example: North’s Hardware

 Transition Matrix

Next Week's Call

Refuses Accepts
This Refuses .35 .65
Week's
Call Accepts .20 .80

© 2003 Thomson/South-Western 52
Slide
 Example: North’s Hardware

 Steady-State Probabilities
Question
How many times per year can Henry expect to
talk to Shirley?
Answer
To find the expected number of accepted calls per
year, find the long-run proportion (probability) of
a call being accepted and multiply it by 52 weeks.

continued
...

© 2003 Thomson/South-Western 53
Slide
 Example: North’s Hardware

 Steady-State Probabilities
Answer (continued)
Let 1 = long run proportion of refused calls
2 = long run proportion of accepted calls
Then,

.35 .65
[    ] = [    ]
.20 .80

continued . . .

© 2003 Thomson/South-Western 54
Slide
 Example: North’s Hardware

 Steady-State Probabilities
Answer (continued)

 +  =  (1)

 +  =  (2)
 +  = 1 (3)

Solve for  and 

continued . . .

© 2003 Thomson/South-Western 55
Slide
 Example: North’s Hardware

 Steady-State Probabilities
Answer (continued)
Solving using equations (2) and (3). (Equation 1 is
redundant.) Substitute  = 1 -  into (2) to give:

.65(1 - 2) +  = 2

This gives  = .76471. Substituting back into

equation (3) gives  = .23529.
Thus the expected number of accepted calls per
year is:
(.76471)(52) = 39.76 or about 40

© 2003 Thomson/South-Western 56
Slide
 Example: North’s Hardware

 State Probability
Question
What is the probability Shirley will accept
Henry's next two calls if she does not accept his call
this week?

© 2003 Thomson/South-Western 57
Slide
 Example: North’s Hardware

 State Probability
Answer
REFUSES
.35 P = .35(.35) = .1225
REFUSES ACCEPTS
.35 P = .35(.65) = .2275
REFUSES .65
REFUSES
.20 P = .65(.20) = .1300
ACCEPTS ACCEPTS
.65 .80 P = .65(.80) = .5200

© 2003 Thomson/South-Western 58
Slide
 Example: North’s Hardware

 State Probability
Question
What is the probability of Shirley accepting
exactly one of Henry's next two calls if she accepts
his call this week?

© 2003 Thomson/South-Western 59
Slide
 Example: North’s Hardware

 State Probability
Answer
The probability of exactly one of the next two
calls being accepted if this week's call is accepted can be
found by adding the probabilities of (accept next week
and refuse the following week) and (refuse next week
and accept the following week) =

.13 + .16 = .29

© 2003 Thomson/South-Western 60
Slide
 Example: Jetair Aerospace

The vice president of personnel at Jetair

Aerospace has noticed that yearly shifts in personnel can
be modeled by a Markov process. The transition matrix
is:

Next Year
Same Pos. Promotion Retire Quit
Fired
Current Year
Same Position .55 .10 .05 .20 .10
Promotion .70 .20 0 .10 0
Retire 0 0 1 0 0
Quit 0 0 0 1 0
Fired 0 0 0 0 1

© 2003 Thomson/South-Western 61
Slide
 Example: Jetair Aerospace

 Transition Matrix

Next Year
Retire Quit Fired Same Promotion
Current Year
Retire 1 0 0 0 0
Quit 0 1 0 0 0
Fired 0 0 1 0 0

Same .05 .20 .10 .55 .10

Promotion 0 .10 0 .70 .20

© 2003 Thomson/South-Western 62
Slide
 Example: Jetair Aerospace

 Fundamental Matrix

-1 -1
1 0 .55 .10 .45 -.10
N = (I - Q ) -1 =  =
0 1 .70 .20 -.70 .80

© 2003 Thomson/South-Western 63
Slide
 Example: Jetair Aerospace

 Fundamental Matrix
The determinant, d = aa - aa
= (.45)(.80) - (-.70)(-.10) = .29
Thus,

.80/.29 .10/.29 2.76 .34

N = =
.70/.29 .45/.29 2.41 1.55

© 2003 Thomson/South-Western 64
Slide
 Example: Jetair Aerospace

 NR Matrix
The probabilities of eventually moving to the
absorbing states from the nonabsorbing states are given
by:

2.76 .34 .05 .20 .10

NR = x
2.41 1.55 0 .10 0

© 2003 Thomson/South-Western 65
Slide
 Example: Jetair Aerospace

 NR Matrix (continued)

Retire Quit Fired

Same .14 .59 .28

NR =
Promotion .12 .64 .24

© 2003 Thomson/South-Western 66
Slide
 Example: Jetair Aerospace

 Absorbing States
Question
What is the probability of someone who was just
promoted eventually retiring? . . . quitting? . . .
being fired?

© 2003 Thomson/South-Western 67
Slide
 Example: Jetair Aerospace

 Absorbing States (continued)

Answer
The answers are given by the bottom row of the
NR matrix. The answers are therefore:

Eventually Retiring = .12

Eventually Quitting = .64
Eventually Being Fired = .24

© 2003 Thomson/South-Western 68
Slide
 End of Chapter 17

© 2003 Thomson/South-Western 69
Slide