2 4

Recap of Last Lecture Utility Theory

• Decision making under risk • Example: the game is to toss a thumbtack. Mark can play
▫ Sensitivity analysis it investigates how our decision might change given a
change in the problem data (e.g., estimation of probability). the game for two times.
▫ What-If Analysis in Excel
• Decision making under uncertainty
▫ Maximax, Maximin, Criterion of realism, Equally likely payoff table
▫ Minimax regret opportunity cost table
• Decision tree
▫ EVSI = EV of conducting survey + cost of survey – EV of not conducting survey
▫ Efficiency of sample information = × 100%
▫ Sensitivity analysis
▫ Estimate probabilities based on prior belief and historical data using Bayes’
Theorem
Risk seeker
Prediction accuracy
• Utility theory
▫ Risk avoider & risk seeker
3
Utility Theory
• After a utility curve has been determined, the utility
Chapter 02 values are used in the decision-making process.
Decision Analysis ▫ Monetary values are replaced with the appropriate utility values.
IIMT3636 ▫ The expected utility for each alternative is computed instead of
Faculty of Business and Economics the EMV.
The University of Hong Kong
Instructor: Dr. Yipu DENG ▫ Then, decision analysis is performed as usual.
 6 8
5) 5000
$500 price S10
.
-
Price UISUO -prices
EU(buy)
-
U (5000
price)
= -
buy EU (buy) 5 U(S200 5 K(3000
bUl F10 5) 3000 price U(3000 price) price) prie)
= 0 x
+ 0
.
- - .
-
. x -
5 $500 utility /500) =
1
5) she 1
10
.
s S10
.
not => EU (noT) = 0
5
.
x / + 0
.
5x0 = 0
. EU (noTS = 0 .
J
not
F(0 5) $0 utility (0) =
0
F10 SS
O O
In-class Exercise: Utility Theory In-class Exercise: Utility Theory
.
.
• Jim is running a risky business that may generate a • Jim is running a risky business that may generate a
profit of $5,000 with probability 0.5 or a zero profit with profit of $5,000 with probability 0.5 or a zero profit with
probability 0.5. loo 3000 400price 2000 probability 0.5. love Love Soro price -
ZULbuy) 0 7 2)
.
0 .
525 0 .
275
EUIbUy) 0 :
9 0 .
75 0 .
55 0 .
0 .
5x 0 .
9 + 0 .
540 .
55
• We know Jim’s utility curve contains the following • We know Jim’s utility curve contains the following
points: U($0)=0, U($1,000)=0.3, U($2,000)=0.55, points: U($0)=0, U($1,000)=0.3, U($2,000)=0.55,
U($3,000)=0.75, U($4,000)=0.9, and U($5,000)=1. U($3,000)=0.75, U($4,000)=0.9, and U($5,000)=1.
• Is Jim a risk-
avoider or a risk seeker? How much can we • How much we can charge him for an insurance that pays
charge him for an insurance wherein we guarantee a $3,000 when the business fails?
$5,000 return if his business is unsuccessful?
5 7
Utility Theory Utility Theory: Selling Insurance
• Example: the game is to toss a thumbtack. Mark can play
the game for two times. Price = Success Failure EU
$2,000
Uninsured $5,000 $0 0.5
Insured $3,000 $3,000 0.75
Price = Success Failure EU
$3,000
Uninsured $5,000 $0 0.5
Insured $2,000 $2,000 0.55
Risk seeker
 10 12
In-class Exercises In-class Exercises
• Maria is considering the possibility of opening a dress • Develop the opportunity loss table.
shop. Her options are to open a small shop, a medium-
sized shop, or no shop at all.
• The market for a dress shop can be good, average, or
bad. The probabilities for these three possibilities are 0.2
for a good market, 0.5 for an average market, and 0.3 for
a bad market.
• What decisions would be made using the minimax regret
criterion and the minimum EOL criterion?
9 11
Utility Theory: Selling Insurance In-class Exercises
• What do you recommend?
Price = Success Failure EU
$1,000
Uninsured $5,000 $0 0.5
Insured $4,000 $2,000 0.725
Price = Success Failure EU
$2,000
Uninsured $5,000 $0 0.5
Insured $3,000 $1,000 0.525
• Calculate the EVPI
EUWPI = 0 .
2 x 100 , 000 + 0 .
5 x
55 , 000 + 0 .
3 x0
=
3) 500 => EUPI 37 500 19
,
500
=
-
, ,
best EMV =
19 , 500
=
18 , or
 3
Introduction
• Linear programming (LP) is a widely used mathematical
Chapter 03 modeling technique designed to help managers in
planning and decision making relative to resource
Linear Programming Models allocation.
IIMT3636
Faculty of Business and Economics • It has been applied in many areas over the past 70 years.
▫ Military, industrial, financial, marketing, accounting, and
The University of Hong Kong agricultural problems
Instructor: Dr. Yipu DENG
13 2
EUSI EU (conduct
survey) of survey
=
+ cost EU (not)
-
900
32 10 42 , 000
900
=
, + ,
000 - =
In-class Exercises Calculate the EVSI
Cost of survey = $10,000
Introduction
First Decision 8 x 80 000 0 2 x 40000 56 Payoffs
000
Second Decision
=
EU 0
- .
,
= .
,
Point Point Favorable Market (0.8)
2 Unfavorable Market (0.2)
$80,000 • Deterministic decision making is without risk and
56 000 –$40,000
uncertainty.
,
34
, 000 Favorable Market (0.8)
Small $50,000
II
Plant
3 Unfavorable Market (0.2)
–$30,000
If
32 900
No Plant
,
–$10,000
1 -
28 , 000
Favorable Market (0.1)
• Many management decisions are deterministic, which
$80,000
-
10 , 000 4 Unfavorable Market (0.9)
–$40,000 involve making the most effective use of an
LI 22 000 Favorable Market (0.1)
organization’s resources.
-
$50,000
,
Small
II
Plant
5 Unfavorable Market (0.9)
–$30,000
No Plant
–$10,000
▫ E.g., Use limited space, money, labor, time, machinery,
warehouse, and raw materials to make products or services
42 , 000
Favorable Market (0.6)
$90,000
42 00 6 Unfavorable Market (0.4)
, –$30,000
28 , 000 Favorable Market (0.6)
Small $60,000
LI
Plant
7 Unfavorable Market (0.4)
–$20,000
I
No Plant
$0
 5 7
Introduction Formulating LP problems
• A typical LP has the following properties and assumptions • Formulating an LP problem involves developing a
▫ Objective function and constraints are linear mathematical model to represent a managerial problem.
All terms used in the objective function and constraints are of the first
degree.
The linear function is related to a sum of terms like 𝛼𝑋, where 𝛼 is a • Steps
constant and X is a decision variable. ▫ Completely understand the managerial problem being faced.
Example: 𝟐𝑨 + 𝟓𝑩 = 𝟏𝟎 linear; ▫ Identify the objective and the constraints.
𝟐𝑨𝟐 + 𝟓𝑩𝟑 + 𝟑𝑨𝑩 = 𝟏𝟎 not linear; ▫ Define the decision variables.
Min{A,B} not linear. ▫ Use the decision variables to write mathematical expressions
▫ Parameters in the objective and constraints (not the decision for the objective function and the constraints.
environment) are certain and do not change during the study
period.
▫ Nonnegative decision variables
4 6
Introduction Quick Check
• A typical LP has the following properties and assumptions. • Are the following objective functions linear?
▫ Maximizing or minimizing one quantity (e.g., profit, cost) ▫ X1 - X2
Example: a manufacturer maximize dollar profits; a railroad distribution
system minimize shipping costs ▫ XY – 3 Yes; No; No
The quantity must be stated clearly and defined mathematically. ▫ X2 + 1/Y
objective function.
▫ One or more constraints that limit the degree to which we can pursue
our objective our objective function is subject to constraints. • Are the following constraints linear?
Example: number of product units to produce & available people and ▫ 2X1 + πX2 ≤ X3
machines; financial portfolio & amount of money available
▫ Alternative courses of action to choose from
▫ X1 ≥ 1/e
Example: a company can produce three different products and the manager ▫ X12 ≤ 4 and X1 ≥ 0 Yes; Yes; Kind of.
can use LP to decide how to allocate resources (1) produce the first
product only; (2) produce equal amounts of each product; (3) produce
products in some other ratio.
No alternatives to choose from no need to use LP
 9 11
Flair Furniture Company Formulating the LP problem
• The company produces inexpensive tables and chairs. • The complete problem is stated mathematically.
• The production process for both products requires a
certain number of labor hours in the carpentry Maximize Profit = $70𝑇 + $50𝐶
department and the painting & varnishing department. Subject to (Amount of resource used ≤ Amount of resource available)
• It wants to determine the best combination of tables and 4𝑇 + 3𝐶 ≤ 240 (carpentry)
chairs to produce to reach the maximum profit. 2𝑇 + 1𝐶 ≤ 100 (painting and varnishing)
HOURS REQUIRED
TO PRODUCE 1 UNIT 𝑇, 𝐶 ≥ 0 (non-negativity)
(T) (C) AVAILABLE HOURS
DEPARTMENT TABLES CHAIRS THIS WEEK
Carpentry 4 3 240
Painting and varnishing 2 1 100
Profit per unit $70 $50
8 10
Product Mix Problem Formulating the LP problem
• One of the most common LP applications • The objective is
Maximize profit
• In such problems, two or more products are produced • The constraints are
using limited resources. 1. The hours of carpentry time used cannot exceed
▫ Objective: maximize profit based on the profit contribution per 240 hours per week
unit of each product
2. The hours of painting and varnishing time used
▫ Constraints: limited resources
cannot exceed 100 hours per week
▫ Decision variables: how many units of each product it should
produce • The decision variables are
T = number of tables to be produced per week
C = number of chairs to be produced per week
 13 15
In-Class Exercise: Trip Planning Graphical Solution Method
• 𝑋 : money spent on flight tickets
• 𝑋 : money spent on hotels • The easiest way to solve a small LP problem is with the
• 𝑋 : money spent on local transportation graphical solution approach.
• 𝑋 : money spent on admission tickets ▫ Only works when there are just two decision variables
• 𝑋 : money spent on food ▫ When there are more than two decision variables, it is
impossible to plot the solution on a 2D graph.
• Maximize utility = 𝑋 + 𝑋 + 𝑋 + 2𝑋 + 2𝑋
• Subject to
• The graphical solution method is invaluable in providing
▫ Total budget constraint: 𝑋 + 𝑋 + 𝑋 + 𝑋 + 𝑋 ≤ 50,000
▫ Safety and time constraint: 𝑋 ≥ 10,000 for 𝑖 = 1,2,3
us with insights into how other approaches work.
▫ Single aspect budget constraint: 𝑋 ≤ 15,000 for all 𝑖
▫ Non-Negativity constraint: 𝑋 ≥ 0 for all 𝑖
12 14
In-Class Exercise: Trip Planning How to Find LP Solutions
• You are planning a family trip to Europe with a total
budget of HKD 50,000. • There are a number of ways to solve an LP.
• Expenses can be categorized into five aspects: (1) flights,
(2) hotels, (3) local transportation, (4) admission tickets,
and (5) food. • Graphical solution method
• The utility generated from admission tickets and food is ▫ Isoprofit line solution method
twice the amount spent on them, while the utility for
flights, hotels, and local transportation is equal to the ▫ Corner point solution method
amount spent.
• To ensure safety and efficiency, you have decided that at
least 20% of the total budget should be allocated to flights, • Excel Solver
hotels, and local transportation each. Additionally, no
single type of expense can exceed 30% of the total budget.
• Your goal is to allocate the budget in a way that maximizes
the total utility.
 17 19
Graphical Solution Method Graphical Solution Method
C
C Equality part of the carpentry constraint
(T, C) points on the graph 100 –
100 –
This Axis Represents –
– (T = 0, C = 80)
the Constraint T ≥ 0 80 –

Number of Chairs
Number of Chairs

80 –
–
– The nonnegativity constraints mean that we are
always working in the first quadrant of a graph. 60 –
60 –
–
– This Axis Represents the
40 –
40 – Constraint C ≥ 0
– (T = 60, C = 0)
–
20 –
20 –
–
–| | | | | | | | | | | | | | | | | | | | | | | |
T 0 20 40 60 80 100 T
0 20 40 60 80 100
Number of Tables Number of Tables
16 18
Graphical Solution Method Graphical Solution Method
• The first step is to identify a set or region of feasible • When Flair produces no tables, the carpentry constraint
solutions by plotting each constraint on a graph and is:
finding the common area. 4(0) + 3𝐶 = 240
3𝐶 = 240 The constraint:
𝐶 = 80 4𝑇 + 3𝐶 ≤ 240
▫ Graph the equality portion of the constraint
Solve for the axis intercepts and draw the line The equality part is a
• Similarly for no chairs: linear function:
4𝑇 + 3(0) = 240 𝐶 = 240 − 4𝑇
▫ Identify the side that satisfies the constraint
4𝑇 = 240
𝑇 = 60
• (0,80) & (60,0)
 21 23
Graphical Solution Method Graphical Solution Method
The Painting and Varnishing Constraint C
C
Our choice of T&C is restricted to this area, and
2𝑇 + 1𝐶 ≤ 100 we need to find a point to maximize the profit.
(T = 0, C = 100) 100 –
100 –
–
–
80 – Painting/Varnishing Constraint

Number of Chairs
80 –
Number of Chairs

–
–
60 –
60 –
– (𝑇 = 30, 𝐶 = 40)
– (70, 40)
40 –
40 –
– Feasible
– (T = 50, C = 0) Carpentry Constraint
20 – Region
20 – (30, 20)
– (50, 5)
–
| | | | | | | | | | | |
| | | | | | | | | | | |
0 20 40 60 80 100 T
0 20 40 60 80 100 T
Number of Tables
Number of Tables
20 22
Graphical Solution Method 4𝑇 + 3𝐶 ≤ 240
Graphical Solution Method
C how can we identify points
that satisfy this constraint? • To produce tables and chairs, both departments must be
100 –
–
Any point on or below the
constraint plot will not
used. So, a solution should satisfy both constraints.
80 – violate the restriction.
Number of Chairs

Any point above the plot

–
will violate the restriction. • Draw a new graph with both constraint plots.
60 –
–
40 –
(30, 40) (70, 40) • The feasible region is where all constraints are satisfied.
– ▫ Any point inside this region is a feasible solution
20 – ▫ Any point outside the region is an infeasible solution
– (30, 20)
| | | | | | | | | | | |
0 20 40 60 80 100 T
Number of Tables
 25 27
Isoprofit Line Solution Method Isoprofit Line Solution Method
C C Profit maximization = go up as much as possible.
At the same time, the isoprofit lines should contain points in
100 – 100 – the feasible region; otherwise, the profit is not achievable.
– –
80 – 80 –
Number of Chairs

Number of Chairs
– –
(0, 42) 4𝑇 + 3𝐶 − 240 = 2𝑇 + 1𝐶 − 100 = 0
60 – 60 – Optimal Solution Point
– $2,100 = $70T + $50C – (T = 30, C = 40)
40 – 40 – Maximum Profit Line
– – $4,100 = $70T + $50C
(30, 0)
20 – 20 –
– –
| | | | | | | | | | | | | | | | | | | | | | | |
0 20 40 60 80 100 T 0 20 40 60 80 100 T
Number of Tables Number of Tables
24 26
Isoprofit Line Solution Method Isoprofit Line Solution Method
C
• How to describe profit in the graph? Recall that 100 – Parallel Isoprofit Lines
profit = $70𝑇 + $50𝐶 – $3,500 = $70T + $50C
80 –

Number of Chairs
– $2,800 = $70T + $50C
• We can let the profit equal some arbitrary but small
60 –
dollar amount such as $2,100. – $2,100 = $70T + $50C
▫ This is a profit level that can be obtained easily 40 –
without violating any constraint. – $4,200 = $70T + $50C
20 – No longer touch the feasible region
–
• It can be achieved by producing 30 tables or 42 chairs or | | | | | | | | | | | |
some combinations of tables and chairs, which is a 0 20 40 60 80 100 T
straight line segment. Number of Tables
The further we move from the origin, the higher our profit will be.
 29 31
Corner Point Solution Method Summary of Graphical Solution Method
C
100 – • For LP problems with two decision variables:
– • Graph all constraints and find the feasible region
(0, 80)
80 – • Isoprofit method
Number of Chairs

– ▫ Select a specific profit line and graph it to find the slope

60 – ▫ Move the objective function line in the direction of increasing profit, while
– maintaining the slope. The last point it touches in the feasible region is the
(30,40) optimal solution.
40 –
▫ Find the values of the decision variables at this last point and compute the profit.
–
20 – • Corner point method
▫ Find the corner points of the feasible region
– (50, 0)
| | | | | | | | | | | | ▫ Compute the profit at each of the feasible corner points
(0, 0) 0 20 40 60 80 100 T ▫ Select the corner point with the best value of the objective function found in the
previous step
Number of Tables
28 30
Corner Point Solution Method Corner Point Solution Method
• Is it possible to have an optimal solution point that is not • Calculate the value of the objective function at each of
a corner point? the corner points
• Observation: the optimal solution must be achieved by • Select the corner point with the best value
one of the corner points. NUMBER OF NUMBER OF PROFIT =
TABLES (T) CHAIRS (C) $70T + $50C
• A second approach is corner point solution method: to 0 0 $0
compare profits achieved by different corner points. 50 0 $3,500
0 80 $4,000
• Remember: This is useful when there are only two 30 40 $4,100
decision variables. There will be too many corner points
when there are too many variables or constraints.
 33 35
Sensitivity Analysis Sensitivity Analysis
• We have found optimal solutions to LP problems under • Two ways to determine how sensitive an optimal
deterministic assumptions.
▫ We assume complete certainty in data and relationships. solution is to changes
▫ e.g., prices are fixed, resource limitations are known, etc.
• However, in the real world, data and relationships are • Trial-and-error approach
dynamic and changing. ▫ This is done with resolving the entire LP problem by computer.
▫ Example: a firm realizes that profit per unit is not $5 as estimated,
but instead is closer to $5.5
▫ Additional resources, such as 10 labor hours or 3 hours of machine • Analytic post-optimality method
time, become available. ▫ After an LP problem has been solved, we attempt to determine a
range of changes in problem parameters that will not affect the
• How can we handle non-deterministic LP problems? optimal solution.
▫ Continue to treat each particular LP problem as a deterministic
situation ▫ This is done without resolving the whole problem.
▫ Utilize sensitivity analysis
32 34
Both the carpentry and P&V constraints are
Slack and Surplus binding (no slack). If we require 𝐶 ≥ 20, then we
have surplus for this constraint (surplus = 20).
Sensitivity Analysis
C
4𝑇 + 3𝐶 ≤ 240 • Difference between the left-hand side and
2𝑇 + 1𝐶 ≤ 100 • Changes in contribution rates for each variable in the
100 – the right-hand side of a constraint: objective function
– • Slack is the amount of a resource that is ▫ What if the profit on product 1 increases by 10%?
(0, 80)
80 – not used.
Number of Chairs

▫ Amount of resource available – amount used

• Changes in the technological coefficients (the numbers in the
–
• Surplus indicates the amount by which a constraint equations)
60 –
greater-than-or-equal-to constraint is ▫ What if new technology will allow a product to be produced in one-
–
(30,40) third the time it used to take?
40 – exceeded.
– ▫ Actual amount - minimum amount • Changes in the available resources (right-hand-side quantities
20 – • If no slack or surplus for the optimal in each constraint)
– (50, 0) solution, then a constraint is binding. ▫ What if less money is available in the advertising budget constraint?
(0, 0) 0
| | | | | | | | | | | | • It can deal not only with errors in estimating input
20 40 60 80 100 T
parameters but also with experiments with possible future
Number of Tables
changes in the firm that may affect profits.
 37 39
Sensitivity Analysis (Technological Coefficients) In-class Exercises
• Changes in needed resources to produce a product • If a new technology is introduced such that one carpentry
the feasible region changes. hour is saved per table, should the company produce more
C C
or fewer tables?
100 – 100 –
– 3𝑇 + 3𝐶 ≤ 240 –
80 – 2𝑇 + 1𝐶 ≤ 100 80 –
4𝑇 + 6𝐶 ≤ 240
–
Maximum Profit Line Produce fewer tables! – Optimal Solution Point
2𝑇 + 1𝐶 ≤ 100
$3,650 = $70T + $50C (T = 20, C = 60)
60 – 60 –
– –
40 – 40 – Maximum Profit Line
– Optimal Solution Point – $4,400 = $70T + $50C
20 – (T = 45, C = 10) 20 –
– –
| | | | | | | | | | | | | | | | | | | | | | | |
0 20 40 60 80 100 T 0 20 40 60 80 100 T
36 38
Sensitivity Analysis (Contribution Rates) Sensitivity Analysis (Resources)
• Changes in objective function • Changes in right-hand-side values of constraints
the slope of the iso-profit line changes. Parallel shifts of boundaries of the feasible region
C C
100 – Maximum Profit Line 100 –
– $3,500 = $70T + $30C –
80 – 80 – Optimal Solution Point
4𝑇 + 3𝐶 ≤ 240 4𝑇 + 3𝐶 ≤ 270
– – (T = 15, C = 70)
2𝑇 + 1𝐶 ≤ 100 Profit = $70T + $51C ? 2𝑇 + 1𝐶 ≤ 100
60 – Maximum Profit = $4,140 60 –
– –
40 – Maximum Profit Line 40 – Maximum Profit Line
– $4,100 = $70T + $50C – $4,550 = $70T + $50C
20 – 20 –
– –
| | | | | | | | | | | | | | | | | | | | | | | |
0 20 40 60 80 100 T 0 20 40 60 80 100 T
 40
In-Class Exercises
• True or False?
False ▫ A new technology that saves one carpentry hour per chair will
lead to the optimal solution of producing fewer chairs.
False ▫ If the number of carpentry hours is increased, it is optimal to
produce more tables.
True ▫ If the profit of a chair is increased to $60, it is optimal to produce
more chairs.
False ▫ If a new technology can save 0.6 carpentry hours per chair, then
the carpentry hour constraint will have positive slack.