Instructors Manual

Chapter 9

102

Chapter 9
I

Chapter Outline

9.1

Formulating a Management Problem as a Discrete Optimization Model

Integer optimization models
Binary optimization models
Mixed-integer optimization models

9.2

Graphical Analysis of Discrete Optimization Models in Two variables

9.3

Computer Solution of Discrete Optimization Problems

General principles on the solvability of a discrete optimization model
Using spreadsheet software to solve a discrete optimization problem

9.4

The Branch-and-Bound Method for Solving a Discrete Optimization Model

Discussion on how this method is used by spreadsheet software to solve
binary optimization models

II

Teaching Tips

1.

It is interesting to show in class how difficult a discrete model can be even for a
small set of decision variables (and even for binary variables). You can show how
the number of feasible solutions grows at an exponential rate according to the
number of variables.

2.

Some of the problems and exercises in this chapter can be also modeled as
network problems. To develop student intuition in this chapter, it helps to present
the alternative formulation in terms of graphs or networks.

3.

The discussion on solving discrete optimization models can be enhanced by

mentioning new optimization techniques such as genetic algorithms, neural
networks, taboo search, etc. In particular, it is not difficult to adapt the capital
budgeting problem (Example 9.2) to be solved by genetic algorithms. Keep in
mind that the most recent versions of Solver include the evolutionary/genetic
search optimization techniques.

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual
III

Chapter 9

103

Answers to Chapter Exercises

9.1
(a) Not shown.
(b) The new optimal strategy is to relocate to London, Madrid, and Hamburg, for
a total cost of $56 million. The optimal fractions of service requests are shown
in the following table (results may differ depending on the tolerance level):

London
Madrid
Paris
Hamburg
Rome

London
1.00
0.00
0.00
0.25
0.00

Fraction of service requests

Madrid
Paris
Hamburg
0.00
0.00
0.00
0.00
0.00
1.00
0.00
0.00
1.00
0.75
0.00
0.00
0.25
0.00
0.75

Rome
0.00
0.00
0.00
0.00
0.00

(c) The results are shown in the table below. Notice that for an average shipping
time across countries of 0.6, the model is no longer feasible (results may differ
depending on the tolerance level).

Average shipping time

0.8
0.7
0.6

London
1
1

Optimal relocation
Madrid
Paris
Hamburg
1
0
1
0
0
1
Infeasible

Rome
0
1

Total cost
$56
$57

(d) In addition to changing the 1.1 value by 1 in the constraint corresponding to

total shipping time, we replace the objective function by this function:
Max
10y1 + 15y2 + 22y3 + 21y4 + 16y5 +
2(1-y1) + 2(1-y2) + 2(1-y3) +
3y4 + 3y5.
The optimal relocation strategy is to close the centers in London and Madrid,
and open a new one in Rome. The total cost of this strategy is $45 millions
(results may differ depending on the tolerance level).
9.2
(a) Consider the following binary optimization model:
Max
25X1 + 40X2 + 125X3 + 100X4 + 60X5 + 15X6.
Subject to:
Average search time:
10X1 + 20X2 + 35X3 + 40X4 + 45X5 + 5X6,
Binary values:
X1, X2, X3, X4, X5, X6 = 0 or 1.
In this model X1 = 1 means that company A is chosen, and X1 = 0 means that
company A is not chosen. Similarly, X2, X3, X4, X5, and X6 indicate
whether or not companies B, C, D, E, and F are to be chosen or not,
respectively.

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual

Chapter 9

104

(b) The results are shown in the table below. As expected, the more the time limit
T increases, the more search engine companies are added to the solution. For
T >= 155 seconds, all of the companies will be in the optimal solution.
T
100
130
150

A
0
1
1

B
1
0
1

C
1
1
1

D
1
1
1

E
0
1
1

F
1
0
0

Hits
280
310
350

9.3
(a) Let Y1, Y2, Y3, Y4, Y5, and Y6 be binary variables each representing
whether or not we place an ATM machine in Arlington, Belmont, Cambridge,
Lexington, Concord, and Winchester, respectively. For instance, Y1 = 1
means that we will place an ATM machine in Arlington. Let Xij represent
whether we establish a connection between cities i and j, for all of the cities.
For instance, X12 = 1 means that a connection between Arlington and
Belmont will be established. Using these variables, we formulate the
following discrete optimization model.
Min Y1 + Y2 + Y3 + Y4 + Y5 + Y6.
Subject to:
Arlington is connected: X11 + X12 + X13 + X14 + X15 + X16 >= 1 Y1,
Belmont is connected:
X21 + X22 + X23 + X24 + X25 + X26 >= 1 Y2,
Cambridge is connected: X31 + X32 + X33 + X34 + X35 + X36 >= 1 Y3,
Lexington is connected: X41 + X42 + X43 + X44 + X45 + X46 >= 1 Y4,
Concord is connected:
X51 + X52 + X53 + X54 + X55 + X56 >= 1 Y5,
Winchester is connected: X61 + X62 + X63 + X64 + X65 + X66 >= 1 Y6,
Arlington distance:
5X12 + 10X13 + 15X14 + 20X15 + 15X16
<= 10(1 Y1),
Belmont distance:
5X21 + 8X23 + 10X24 + 15X25 + 12X26
<= 10(1 Y2),
Cambridge distance:
10X31 + 8X32 + 15X34 + 20X35 + 10X36
<= 10(1 Y3),
Lexington distance:
15X41 + 10X42 + 15X43 + 10X45 + 12X46
<= 10(1 Y4),
Concord distance:
20X51 + 15X52 + 20X53 + 10X54 + 12X56
<= 10(1 Y5),
Winchester distance:
15X61 + 12X62 + 10X63 + 12X64 + 12X65
<= 10(1 Y6),
Logical constraints:
Yj >= Xij, for all i, j = 1, , 6,
Binary variables
Yi, Xj = 0 or 1, for all i, j = 1, , 6.
(b) An optimal solution is to place two ATM machines, one in Cambridge and the
other in Concord. Another solution is to place the ATM machines in
Cambridge and Lexington.

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual

Chapter 9

105

9.4
(a) Let X1, , X9 denote whether or not we choose each of the TV shows. For
instance, X1 = 1 means that we choose Cheers. In addition, let T denote
whether or not we schedule more than 3 violence shows. Using these
variables, we obtain the following discrete optimization model.
Max 6X1 + 10X2 + 9X3 + 4X4 + 5X5 + 2X6 + 6X7 + 7X8 + 8X9 4T.
Subject to:
Total shows:
X1 + X2 + X3 + X4 + X5 + X6 + X7 + X8 + X9 = 5,
Public=violence: X3 + X6 + X7 + X9 = X2 + X3 + X4 + X6,
Jake+LALaw:
X3 + X4 >= X7,
Dry shows:
X7 + X9 <= 1,
Comedy+drama: 2(X1 + X2 + X3 + X4 + X7) + 1 >= X1 + X5 + X6 + X8,
Violence cost:
T + 3 >= X2 + X3 + X4 + X6,
Binary:
X1, X2, X3, X4, X5, X6, X7, X8, X9, T = 0 or 1.
(b) The optimal solution is to show Cheers, Dynasty, LA Law, Beaches, and
Urban Action for Education, with total revenues of $40 millions.
9.5
(a) Let X1, X2, X3, and X4 represent the amount to be ordered from each of the
four furniture companies, respectively. Let Y1, Y2, Y3, and Y4 be binary
variables representing whether or not we place an order with each of the four
furniture companies, respectively. For instance, Y1 = 1 means that we place
an order with Carolina Woodworks. Using these variables, we obtain the
following discrete optimization model.
Min
2,500X1 + 2,450X2 + 2,510X3 + 2,470X4 +
10,000Y1 + 20,000Y2 + 13,000Y4.
Subject to:
Required sets:
X1 + X2 + X3 + X4 = 2,000,
CW bounds:
Y1 <= X1 <= 1,000Y1,
NM bounds:
Y2 <= X2 <= 1,000Y2,
AFD bounds:
Y3 <= X3 <= 1,000Y3,
LAC bounds:
Y4 <= X4 <= 1,000Y4,
Integer:
X1, X2, X3, X4 >= 0 and integer,
Binary:
Y1, Y2, Y3, Y4 = 0 or 1.
(b) The optimal solution is to order 1,200 furniture sets from Nashawtuc
Millworks and 800 furniture sets from Lancaster Artisan Company. The cost
of this solution is $4,949,000 (results may differ depending on the tolerance
level).
(c) We add four more variables: X5 and X6 to represent the furniture sets ordered
from Delaware Mills and Y1 and Y2 representing whether or not we place an
order with Delaware Mills. X5 represents an order between up to 1,000 sets
and X6 represents an order between 1,000 and 1,500 sets. Using these
variables, we obtain the following reviewed discrete optimization model.
Min 2,500X1 + 2,450X2 + 2,510X3 + 2,470X4 + 2,530X5 + 2,430X6 +
10,000Y1 + 20,000Y2 + 13,000Y4 + 9,000Y5 + (7,000 + 100x1,000)Y6.

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual

Chapter 9

106

Subject to:
Required sets:
X1 + X2 + X3 + X4 + X5 + X6= 2,000,
CW bounds:
Y1 <= X1 <= 1,000Y1,
NM bounds:
Y2 <= X2 <= 1,000Y2,
AFD bounds:
Y3 <= X3 <= 1,000Y3,
LAC bounds:
Y4 <= X4 <= 1,000Y4,
DM1 bounds:
Y5 <= X5 <= 1,000Y5,
DM2 bounds:
1,000Y6 <= X6 <= 1,500Y6,
Logical:
Y5 + Y6 <= 1,
Integer:
X1, X2, X3, X4, X5, X6 >= 0 and integer,
Binary:
Y1, Y2, Y3, Y4, Y5, Y6 = 0 or 1.
The optimal solution is the same as before.
9.6

Let A1, B1, C1, D1, E1, A2, B2, C2, D2, E2, A3, B3, C3, D3, and E3 represent
the amounts to be ordered for each module and engine for each year, respectively.
Using these variables, we obtain the following discrete optimization problem:
Min
0.7A1 + 2.5B1 + 6.0C1 + 1.3D1 + 9.0E1 +
0.6A2 + 2.2B2 + 5.5C2 + 1.1D2 + 8.5E2 +
0.5A3 + 2.0B3 + 5.0C3 + 1.0D3 + 7.8E3
Subject to:
Inventory A1: A1 + E1 >= 5,
Inventory B1: B1 + E1 >= 4,
Inventory C1: C1 + E1 >= 4,
Inventory D1: D1 + E1 >= 2,
Inventory E1: E1 >= 1,
Inventory A2: A1 + A2 + E1 + E2 >= 8,
Inventory B2: B1 + B2 + E1 + E2 >= 6,
Inventory C2: C1 + C2 + E1 + E2 >= 6,
Inventory D2: D1 + D2 + E1 + E2 >= 10,
Inventory E2: E1 + E2 >= 1,
Inventory A3: A1 + A2 + A3 + E1 + E2 + E3 >= 13,
Inventory B3: B1 + B2 + B3 + E1 + E2 + E3 >= 12,
Inventory C3: C1 + C2 + C3 + E1 + E2 + E3 >= 11,
Inventory D3: D1 + D2 + D3 + E1 + E2 + E3 >= 12,
Inventory E3: E1 + E2 + E3 >= 3,
Integer:
A1, B1, C1, D1, E1,
A2, B2, C2, D2, E2,
A3, B3, C3, D3, E3 >= 0 and integer.
The optimal schedule is shown in the following table. The cost of this schedule is
$98.9 millions.

Year 1
Year 2
Year 3

Engine modules and engines

A
B
C
1
0
0
0
0
0
1
1
0

D
0
1
0

E
5
4
2

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual

Chapter 9

107

9.7
(a) Let W1, , W12 be the amounts to be produced of the Warrior model for each
month. Let C1, , C12 be the amounts to be produced of the Chomper model
for each month. To control the inventory levels of the Warrior and Chomper
models, respectively, we use the variables IWi and ICi, for i =1, , 12. To
determine what to produce each month, we use the binary variables Xi and Yi,
for i =1, , 12. For instance Xi = 1 indicates that we will produce the Warrior
model. Finally, we use the variables Si, for i =1, , 12, to indicate whether or
not we switch production in a given month. We put together all of these
variables in the following discrete optimization model.
Min
7000(W1 + + W6) + 7700(W7 + + W12) +
6000(C1 + + C6) + 6600(C7 + + C12) +
700(IW1 + + IW12) + 600(IC1 + + IC12) +
1500(S1 + + S12).
Subject to:
Warrior inventory:
Wi + IWi-1 IWi = DWi, i = 1, , 12,
Chomper inventory:
Ci + ICi-1 ICi = DCi, i = 1, , 12,
Either model:
Xi + Yi <= 1, i = 1, , 12,
Warrior capacity:
Xi <= Wi <= 100Xi, i = 1, , 12,
Chomper capacity:
Yi <= Ci <= 110Yi, i = 1, , 12,
Switching 1:
Xi-1 + Yi Si <= 1, i = 1, , 12,
Switching 2:
Yi-1 + Xi Si <= 1, i = 1, , 12,
Binary:
Xi, Yi, Si = 0 or 1, i = 1, , 12,
Integer:
Wi, Ci, IWi, ICi >= 0 and integer, i = 1, , 12.
In this model we have IW0 =IC0 = 50, X0 = 1, and Y0 = 0. The variables DWi
and DCi represent the demand for the Warrior and Chomper models,
respectively, for each month.
(b) The optimal production schedule is shown in the table below. The total cost of
this schedule is $6,283,800,000 (results may differ depending on the tolerance
level).

Month
January
February
March
April
May
June
July
August
September
October
November
December

Production Schedule
W
C
23
0
0
83
71
0
0
74
63
0
0
85
73
0
0
85
78
0
0
95
81
0
0
46

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual

Chapter 9

108

9.8
(a) Let F1, , F20 be binary variables indicating whether or not Linda takes any
of the 20 courses during the fall term. For instance, F1 = 1 means that Linda
takes course number 1 in the fall. Similarly, let S1, , S20 be binary variables
indicating whether or not Linda takes any of the 20 courses during the spring
term. Let OF1, , OF20 and OS1, .., OS20 be binary constants indicating
whether each of the classes is offered in the fall and spring terms,
respectively. Finally, let I1, .., I20 be the interest level assigned by Linda to
each of the 20 courses. Consider the following discrete optimization model.
Max
I1(F1 + S1) + + I20(F20 + S20).
Subject to:
Fall course offering:
F1 <= OF1, , F20 <= OF20,
Spring course offering: S1 <= OS1, , S20 <= OS20,
Fall term maximum:
F1 + + F20 <= 5,
Spring term maximum: S1 + + S20 <= 5,
Prerequisites course 5: S5 <= F4,
Prerequisites course 7: S7 <= F2, S7 <= F6,
Prerequisites course 8: S8 <= F1, S8 <= F3 + S3,
Prerequisites course 9: S9 <= F1,
Prerequisites course 11: S11 <= F1, S11 <= F10,
Prerequisites course 13: S13 <= S9, S13 <= F12,
Prerequisites course 14: S14 <= F3 + S3,
Prerequisites course 16: S16 <= F15,
Prerequisites course 17: S17 <= F4,
Prerequisites course 18: S18 <= F10, S18 <= F12, S18 <= S17,
Prerequisites course 19: F19 <= F4,
Fall term requirement:
F1 + F2 + F3 + F6 + F20 >= 1,
Financial or options:
S14 <= (1 S8),
Marketing:
F12 + S13 >= 1,
O.M.:
F10 + S11 >= 1,
Binary:
F1, , F20, S1, , S20 = 0 or 1.
The optimal course schedule is to take courses 1, 2, 10, 12, and 20 during the
fall term and courses 3, 8 , 9, 11, and 13 during the spring term. Linda would
get 42 points of interest level by taking these courses.
(b) The other optimal schedule is to take courses 1, 2, 10, 12, and 20 during the
fall term and courses 3, 9, 11, 13, and 14 during the spring term. Linda would
also get 42 points of interest level by taking these courses.
9.9
(a) Let X1, , X10 be as defined in this problem and let Y1, , Y20 be binary
variables indicating whether or not we include a company in the portfolio. For
instance, Y1 = 1 indicates that we will use company 1 in the portfolio. Using
these variables, we obtain the following nonlinear discrete optimization
model.

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual
Min

Chapter 9

109

|X1 0.15| + | X2 0.05| + | X3 0.08| + | X4 0.11| +

| X5 0.14| + | X6 0.09| + | X7 0.08| + | X8 0.08| +
| X9 0.14| + |X10 0.08|.

Subject to:
Number of companies:
Fractions:
Investment:
Technology:
Last month fractions:

Y1 + + Y10 = 6,
X1 + + X10 = 1,
X1 <= Y1, , X10 <= Y10,
0.45 <= X1 + X2 + X5 + X7 + X10 <= 0.55,
|X1 0.12| + | X2 0.08| + | X3 0.07| +
| X4 0.14| + | X5 0.17| + | X6 0.06| +
| X7 0.09| + | X8 0.04| + | X9 0.18| +
|X10 0.05| <= 0.60,
Binary:
Y1, , Y10 = 0 or 1,
Non-negativity
X1, , X10 >= 0.
(b) There might be more than one optimal portfolio. The following table shows
two of them. For both portfolios the optimal difference is 0.58 (results may
differ depending on the tolerance level).

Company
Selected
Fraction

Portfolio 1
1
2
1
0
0.28 0.00

3
0
0.00

4
1
0.13

5
1
0.17

6
1
0.11

7
0
0.00

8
1
0.10

9
1
0.20

10
0
0.00

Company
Selected
Fraction

Portfolio 2
1
2
1
0
0.38 0.00

3
1
0.08

4
1
0.11

5
1
0.17

6
1
0.09

7
0
0.00

8
0
0.00

9
1
0.17

10
0
0.00

IV

Answers to Chapter Cases

INTERNATIONAL INDUSTRIES, INC.

(a) We use the binary variables Ii, Mi, and Si to indicate whether or not to invest,
maintain, or sell division I, respectively, for i = 1, , 50. We represent the net
present values for each decision (invest, maintain, or sell) and for each
division as NIi, NMi, and NSi, respectively, for i = 1, , 50. Similarly, we
represent the investments required for each decision (invest, maintain, or sell)
and for each division as IIi, IMi, and ISi, respectively, for i = 1, , 50. Finally,
we represent the cash flow for next year for each decision (invest, maintain, or
sell) and for each division as CIi, CMi, and CSi, respectively, for i = 1, , 50.
Using these variables, we obtain the following discrete optimization model.

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual

Chapter 9

110

Max NI1I1 + NM1M1 + NS1S1 + + NI50I50 + NM50M50 + NS50S50.

Subject to:
Cash-flow: II1I1 + IM1M1 + IS1S1 + + II50I50 + IM50M50 + IS50S50
<=
CI1I1 + CM1M1 + CS1S1 + + CI50I50 + CM50M50 + CS50S50
Constraint (a):
S1 = I2, S1 = S4, S1 = S3,
Constraint (b):
I6 = I7,
Constraint (c):
S3 + S4 + S5 + M6 + S7 + S9 <= 1,
Constraint (d):
S2 + I4 + M6 + S6 + S8 + S12 + I14 <= 1,
Constraint (e):
I24 <= I28,
Constraint (f):
M30 <= I32 + M32,
Make a decision: Ii + Mi + Si = 1, i = 1, , 50,
Binary:
I1, , I50, M1, , M50, S1, , S50 = 0 or 1.
The optimal set of choices for each division is shown in the table below. The
net present value of optimal choices is $33,689 millions.
Division
Number Choice
1
Maintain
2
Sell
3
Invest
4
Maintain
5
Maintain
6
Invest
7
Invest
8
Maintain
9
Maintain
10
Sell
11
Sell
12
Maintain
13
Invest
14
Sell
15
Maintain
16
Maintain
17
Invest
18
Invest
19
Invest
20
Sell
21
Sell
22
Invest
23
Maintain
24
Sell
25
Sell
26
Invest
27
Invest
28
Sell
29
Invest
30
Invest
31
Sell
32
Sell
33
Sell

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual

Chapter 9
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50

111

Sell
Maintain
Sell
Invest
Invest
Maintain
Invest
Sell
Invest
Maintain
Invest
Invest
Maintain
Invest
Sell
Maintain
Sell

(b) If constraints (a) to (f) are removed, then the optimal NPV will increase to
$39,278 millions. This represents a penalty of $5,589 millions for imposing
the restrictions.
(c) To find the second best and the third best solutions you can impose an
additional constraint on the total NPV. For instance, we know that the optimal
NPV is $33,689 millions. If we impose the constraint NI1I1 + NM1M1 + NS1S1
+ + NI50I50 + NM50M50 + NS50S50 <= $32,000 (or any other bound close to
the optimal objective value), we will force the computer to find a solution
with an objective value close to the original optimal solution. By decreasing
even more the right-hand side of this constraint and re-solving the model, we
can find other good solutions.

SUPPLY CHAIN MANAGEMENT AT DELLMAR, INC.

(a) To describe our model, we index the months as follows: 0 corresponds to
March, 1 corresponds to April, and 2 corresponds to May. We also use the
following variables and parameters:
HOi: Amount produced at New Hampshire and shipped to Ohio,
for months i = 1, 2.
HJi: Amount produced at New Hampshire and shipped to New Jersey,
for months i = 1, 2.
OWi: Amount shipped from Ohio to West Coast, for months i = 1, 2.
OMi: Amount shipped from Ohio to Midwest, for months i = 1, 2.
JMi: Amount shipped from New Jersey to Midwest, for months i = 1, 2.
JEi: Amount shipped from New Jersey to East Coast, for months i = 1, 2.
IOi: Inventory at Ohio center, for months i = 0, 1, 2.
IJi: Inventory at New Jersey center, for months i = 0, 1, 2.
IWi: Inventory at West Coast center, for months i = 0, 1, 2.
IMi: Inventory at Midwest center, for months i = 0, 1, 2.
Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual

Chapter 9

112

IEi: Inventory at East Coast center, for months i = 0, 1, 2.

THOi: Whether or not to ship from New Hampshire to Ohio,
for months i = 1, 2.
THJi: Whether or not to ship from New Hampshire to New Jersey,
for months i = 1, 2.
TOWi: Whether or not to ship from Ohio to West Coast,
for months i = 1, 2.
TOMi: Whether or not to ship from Ohio to Midwest,
for months i = 1, 2.
TJMi: Whether or not to ship from New Jersey to Midwest,
for months i = 1, 2.
TJEi: Whether or not to ship from New Jersey to East Coast,
for months i = 1, 2.
DWi: Demand at West Coast center, for months i = 1, 2.
DMi: Demand at Midwest center, for months i = 1, 2.
DEi: Demand at East Coast center, for months i = 1, 2.
Using these variables, we have the following discrete optimization model.
Min 10(OW1 + OW2 + OM1 + OM2 + JM1 + JM2 + JE1 + JE2 +
HO1 + HO2 + HJ1 + HJ2) +
5000(THO1 + THO2) + 4000(THJ1 + THJ2) + 4000(TOW1 + TOW2) +
3000(TOM1 + TOM2) + 5000(TJM1 + TJM2) + 3000(TJE1 + TJE2) +
5(IO1 + IO2 + IJ1 + IJ2) + 10(IW1 + IW2 + IM1 + IM2 + IE1 + IE2).
Subject to:
Inventory Ohio:
HOi + IOi-1 IOi = OWi + OMi, i = 1,2,
Inventory New Jersey: HJi + IJi-1 IJi = JWi + JMi, i = 1,2,
Inventory West Coast:
OWi + IWi-1 IWi = DWi, i = 1,2,
Inventory Midwest:
OMi + JMi + IMi-1 IMi = DMi, i = 1,2,
Inventory East Coast:
JEi + IEi-1 IEi = DEi, i = 1,2,
Production capacity:
HOi + HJi <= 50,000, i = 1,2,
Shipping decision 1:
THOi <= HOi <= 50,000 THOi, i = 1,2,
Shipping decision 2:
THJi <= HJi <= 50,000 THJi, i = 1,2,
Shipping decision 3:
TOWi <= OWi <= 50,000 TOWi, i = 1,2,
Shipping decision 4:
TOMi <= OMi <= 50,000 TOMi, i = 1,2,
Shipping decision 5:
TJMi <= JMi <= 50,000 TJMi, i = 1,2,
Shipping decision 6:
TJEi <= JEi <= 50,000 TJEi, i = 1,2,
Binary:
THOi,THJi,TOWi,TOMi,TJMi,TJEi = 0 or 1, i = 1,2,
Integer:
HOi,HJi,OWi,OMi,JMi,JEi,IOi,IJi,IWi,IMi,IEi >= 0
and integer, i = 1,2.
(b) If optimized, the average monthly cost in April and May is $1,230,000. The
optimal production and shipping schedule, as well as the optimal inventory
levels are shown in the tables below.

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual

Chapter 9

113

Month
1
2

Production and
Shipping
HO
HJ
32000
18000
0
50000

OW
18000
20000

OM
14000
0

JM
0
25000

Month
0
1
2

Inventory
IO
20000
20000
0

IW
2000
0
0

IM
1000
0
0

IE
2000
0
0

IJ
10000
5000
0

JE
23000
30000

(c) If production is increased by 10%, then the average monthly cost decreases to
$1,217,500 per month by using the optimal solution of the revised model. This
represents savings of $12,500 per month.
(d) The average monthly costs for each scenario are shown in the table below. NF
means nor feasible.

Current capacity
10% increase

Minimum Inventory
Level
0
500
1000
$1,230,000
NF
NF
$1,217,500 $1,257,500 $1,297,500

(e) Based on the results from the previous items, it is not recommendable to
increase capacity by 10% while keeping minimum inventory levels at zero.
This is because the cost of such extra capacity exceeds the amount of savings
obtained by the additional capacity ($55,000 > $12,500). If minimum
inventory levels of 500 or 1000 units per distribution center are necessary,
then production needs to increase by 10%, otherwise it will be impossible to
satisfy the minimum level requirements.

THE NATIONAL BASKETBALL DREAM TEAM

(a) Let X1, , X20 represent whether or not we choose any of the players. For
instance, X1 = 1 means that we choose player 1 to be part of the team.
Similarly, we use the parameters Pi, Ri, Ai, Di, and Hi to denote the points,
rebounds, assists, defensive ability, and height of the i-th player, for i = 1, ,
20. Using these variables and parameters, we obtain the following discrete
optimization model.

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.

Instructors Manual
Max
Subject to:
Points:
Rebounds:
Assists:
Defense:
Height:
Play makers:
Shooting guards:
Forwards:
Centers:
NCAA:
Logical 5 & 9:
Logical 2 and 19:
Bulls:
Binary:

Chapter 9

114

P1X1 + + P20X20,
P1X1 + + P20X20 >= 216,
R1X1 + + R20X20 >= 84,
A1X1 + + A20X20 >= 72,
D1X1 + + D20X20 >= 102,
H1X1 + + H20X20 >= 948,
X1 + + X5 >= 3,
X4 + + X11 >= 4,
X9 + + X16 >= 4,
X16 + + X20 >= 3,
X4 + X8 + X15 + X20 >= 2,
X5 <= 1 X9,
X2 = X19,
X1 + X7 + X12 + X16 <= 3,
X1, , X20 = 0 or 1.

The optimal solution is to choose players 2, 3, 4, 5, 7, 10, 12, 13, 15, 17, 18,
and 19. The average points per game among the players of this team are 22.
(b) There is no feasible solution under the new set of constraints.
(c) By re-solving the model for 14 players instead of just 12, we obtain an
optimal team with almost the same objective value as before. This team
contains the players 2, 3, 4, 5, 6, 7, 10, 11, 12, 13, 15, 17, 18, and 19.
Therefore, the alternate players might be 6 and 11.

Manual to accompany Data, Models & Decisions: The Fundamentals of Management Science by Bertsimas and Freund. Copyright
2000, South-Western College Publishing. Prepared by Manuel Nunez, Chapman University.