0% found this document useful (0 votes)
28 views

Lecture 9

The document discusses sensitivity analysis, which examines how changes to parameters in a linear programming model impact the optimal solution. It provides an overview of sensitivity analysis, how to read sensitivity reports in Excel, and examples of analyzing changes to objective function coefficients and right-hand side constraint values. The sensitivity report allows users to determine if the optimal solution remains the same or changes when parameters are adjusted.

Uploaded by

yeunghc519
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
28 views

Lecture 9

The document discusses sensitivity analysis, which examines how changes to parameters in a linear programming model impact the optimal solution. It provides an overview of sensitivity analysis, how to read sensitivity reports in Excel, and examples of analyzing changes to objective function coefficients and right-hand side constraint values. The sensitivity report allows users to determine if the optimal solution remains the same or changes when parameters are adjusted.

Uploaded by

yeunghc519
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 33

LECTURE 9

LINEAR PROGRAMMING
SENSITIVITY ANALYSIS

1
AGENDA

 Sensitivity Analysis: How to read sensitivity report in Excel


 Summary

2
INTRODUCTION TO SENSITIVITY ANALYSIS

 Not only is the optimal solution to a linear program


formulated for a problem essential, but the extent of impact
on the optimality is also informative when only one
parameter of the problem changes.

 A more explicit statement for the latter is that you want to


know how sensitive the optimal solution is, when only one
coefficient in the objective function changes or only one
value on the rhs of a constraint changes.

 Sensitivity analysis allows you to immediately know whether


the solution remains optimal, and:

 Understand trade-offs.

 Make quick adjustments (e.g. re-solve).

 Obtain intuition.

3
WHAT IS SENSITIVITY ANALYSIS

 The study of how changes in the parameters of a linear


program affect the optimal solution ( 𝑥 ∗ ) and the
optimal value of the objective function.

Objective function: change


in coefficients

𝐌𝐀𝐗 𝐌𝐈𝐍 𝑐 𝑥 𝑐 𝑥 𝑐 𝑥 … 𝑐 𝑥
s.t.
𝐴 𝑥 𝐴 𝑥 𝐴 𝑥 … 𝐴 𝑥 𝐵
Constraints:
𝐴 𝑥 𝐴 𝑥 𝐴 𝑥 … 𝐴 𝑥 𝐵
change in
𝐴 𝑥 𝐴 𝑥 𝐴 𝑥 … 𝐴 𝑥 𝐵 values on
… the rhs

𝐴 𝑥 𝐴 𝑥 𝐴 𝑥 … 𝐴 𝑥 𝐵
𝑥 0 or unrestricted for each 𝑖 1, … , 𝑛

4
HOW TO READ SENSITIVITY REPORT

 Example: Par Inc.

 Objective function: Change in only one coefficient

 “Allowable increase” and “allowable decrease”

 “Reduced cost”

 Constraints: Change in only one value on the right-hand-


side (rhs)

 “Shadow price”

 “Allowable increase” and “allowable decrease”

 Simultaneous changes

 “100% rule”

5
SENSITIVITY REPORT IN EXCEL

 Example: Par Inc.


Profit Contribution 𝐌𝐀𝐗 10 𝑥 9𝑥
s.t.
Cutting & Dyeing: 𝑥 𝑥 630
Sewing: 𝑥 𝑥 600
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135
Non-negativity: 𝑥 0, 𝑥 0

 Optimal solution: 𝑥 540, 𝑥 252

 Optimal profit: 10 540 9 252 7668

 Sensitivity report:

Obj
function
analysis

Constraints
analysis
CHANGE IN A COEFFICIENT

 Q1: What if the profit of a deluxe bag is $10, instead of $9? What
will the production plan and profit become?
Profit Contribution 𝐌𝐀𝐗 10 𝑥 9𝑥 10𝑥 10𝑥
s.t.
Cutting & Dyeing: 𝑥 𝑥 630
Sewing: 𝑥 𝑥 600
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135
Non-negativity: 𝑥 0, 𝑥 0

7
CHANGE IN A COEFFICIENT

 Q1: What if the profit of a deluxe bag is $10, instead of $9? What
will the production plan and profit become?
Profit Contribution 𝐌𝐀𝐗 10 𝑥 9𝑥 10𝑥 10𝑥
s.t.
Cutting & Dyeing: 𝑥 𝑥 630
Sewing: 𝑥 𝑥 600
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135
Non-negativity: 𝑥 0, 𝑥 0

 𝑥 : Actual increase 10 9 5.2857 allowable increase.

 Hence, the optimal solution 540,252 does not change.


8

 New profit = old profit + 10 9 252 7668 252 7920


“ALLOWABLE INCREASE” AND “ALLOWABLE
DECREASE” IN THE “ADJUSTABLE CELLS”

 Suppose coefficient on 𝑥 changes from 𝑐 , →𝑐, .

 Let Δ𝑐 𝑐, -𝑐, . (previous e.g. 𝛥 𝑐 10 9 1 .

 The “allowable” range of variable 𝑥 tells you how much


its coefficient in the objective function can be increased or
decreased without changing the optimal solution (i.e. the
optimal values of decision variables).

 Requires all other parameters to stay the same.

 New value of the objective function (𝑶𝑩𝑱𝒏𝒆𝒘 ):

 Within “allowable” range of 𝑥 , we have the same optimal


solution 𝑥 ∗ 𝑥 ∗ , and
𝑂𝐵𝐽 𝑂𝐵𝐽 𝛥𝑐 𝑥 ∗

(previous e.g. 𝑂𝐵𝐽 7,668 1 252 .

 Outside the “Allowable” range, we have a different optimal


solution (𝑥 ∗ 𝑥 ∗ ), so we re-formulate and resolve!
9
DETECT MULTIPLE OPTIMAL SOLUTIONS

 Question: Is the optimal solution unique?

 Answer:

 Check the allowable increase and allowable decrease of


variable cells in the sensitivity report.

 If there exists a zero in either allowable increase or


allowable decrease, then there are multiple optimal
solutions. Otherwise, the optimal solution is unique.

10
DETECT MULTIPLE OPTIMAL SOLUTIONS

 Par Example: Suppose the profit of a standard bag is now $12, and
the profit of a deluxe bag is $8.
Profit Contribution 𝐌𝐀𝐗 12 𝑥 8𝑥
s.t.
Cutting & Dyeing: 𝑥 𝑥 630
Sewing: 𝑥 𝑥 600
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135
Non-negativity: 𝑥 0, 𝑥 0

11
GRAPHICAL EXPLANATION:
MULTIPLE OPTIMAL SOLUTIONS
Cutting & Dyeing: 𝑥 𝑥 630
Sewing: 𝑥 𝑥 600
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135
Opt
solution
identified

 There are infinitely many optimal solutions, where the objective line
touches the feasible region by overlapping with the finishing
constraint line.
12

 Points on the overlapping segment yields the same optimal profit.


GRAPHICAL EXPLANATION:
MULTIPLE OPTIMAL SOLUTIONS
Cutting & Dyeing: 𝑥 𝑥 630
Sewing: 𝑥 𝑥 600
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135

Shift in profit
line if profit of
𝑥 decreases
or profit of 𝑥
increases.

 If the slope of the objective line (profit) changes, the LP has to be re-
solved to obtain a new optimal solution.This happens when:

 Unit profit of 𝑥 decreases (allowable decrease = 0)


13

 Unit profit of 𝑥 increases (allowable increase = 0).


CHANGE IN OBJECTIVE FUNCTION COEFFICIENT –
REDUCED COST
 Suppose we have the following golf bag problem:
Profit Contribution 𝐌𝐀𝐗 10 𝑥 6𝑥
s.t.
Cutting & Dyeing: 𝑥 𝑥 630 Optimal
solution:
Sewing: 𝑥 𝑥 600
708, 0
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135
Non-negativity: 𝑥 0, 𝑥 0
 Q2: What is the minimal profit contribution of a deluxe bag so that
it is worthwhile to be produced?

14
CHANGE IN OBJECTIVE FUNCTION COEFFICIENT –
REDUCED COST
 Suppose we have the following golf bag problem:
Profit Contribution 𝐌𝐀𝐗 10 𝑥 6𝑥
s.t.
Cutting & Dyeing: 𝑥 𝑥 630 Optimal
solution:
Sewing: 𝑥 𝑥 600
708, 0
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135
Non-negativity: 𝑥 0, 𝑥 0
 Q2: What is the minimal profit contribution of a deluxe bag so that
it is worthwhile to be produced?

 Reduced Cost of 𝑥 tells us how much 𝑐 (coefficient in objective


function) needs to change in order to move to a new optimal
solution with 𝑥 ∗ 0. 15

 We need to reduce the profit of the deluxe bag by -0.67 = Increase


the price by $0.67.
CHANGE IN OBJECTIVE FUNCTION COEFFICIENT –
REDUCED COST
Cutting & Dyeing: 𝑥 𝑥 630
Sewing: 𝑥 𝑥 600
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135

Shift in profit
𝑃𝑟𝑜𝑓𝑖𝑡
10 𝑥 6𝑥 line in order for
deluxe bags to
be produced in
an optimal
solution.

 If slope of the objective line (profit) = slope of the finishing constraint


line, then infinitely many optimal solutions.

 We resolve to get an optimal solution with 𝑥 0.


16
 We can find this change Δ in the unit profit of 𝑥 as follows:

10 1 2
⇒Δ 0.6666 … Reduced cost!
6 Δ 2/3 3
CHANGE IN RHS VALUE OF A CONSTRAINT –
SHADOW PRICE

 Q3: Suppose we increased our production capacity and can perform


an additional 40 hours of cutting & dyeing. What will happen to the
optimal production plan and optimal profit?
Profit Contribution 𝐌𝐀𝐗 10 𝑥 9𝑥
s.t.
Cutting & Dyeing: 𝑥 𝑥 630 670
Sewing: 𝑥 𝑥 600
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135
Non-negativity: 𝑥 0, 𝑥 0

1000

800
Feasible region
# Deluxe Bags

600
becomes bigger.

400

200

0
0 100 200 300 400 500 600 700 80017
# Standard Bags

old cut sew finish


inspect profit new cut
CHANGE IN RHS VALUE OF A CONSTRAINT –
SHADOW PRICE

 Q3: Suppose we increased our production capacity and can perform


an additional 40 hours of cutting & dyeing. What will happen to the
optimal production plan and optimal profit?
Profit Contribution 𝐌𝐀𝐗 10 𝑥 9𝑥
s.t.
Cutting & Dyeing: 𝑥 𝑥 630 670
Sewing: 𝑥 𝑥 600
Finishing: 𝑥 𝑥 708
Inspection & Packaging: 𝑥 𝑥 135
Non-negativity: 𝑥 0, 𝑥 0

 1st constraint: Actual increase = 40 52.36 = allowable increase.

 The optimal solution (production plan) may change, but we can use
the shadow price to compute the new optimal profit. 18

 New profit = old profit + shadow price x rhs value change


7668 4.375 670 630 7843
INTERPRETATIONS OF SHADOW PRICE

 Shadow price:

 The marginal value of a resource

 The change to the objective function value, assuming that


only the corresponding resource increases by 1 unit.

 Focus on “binding” constraint’s RHS (why?)

 Look at the “Allowable Increase” and “Allowable Decrease”

 Suppose the RHS of the 𝑗 -th constraint changes from


𝐵, →𝐵, . Let Δ𝐵 𝐵, 𝐵, .

(previous e.g. 𝛥𝐵 𝐵 , 𝐵 , 670 630 40 .

 Within “allowable” range, :

𝑂𝐵𝐽 𝑂𝐵𝐽 𝑠ℎ𝑎𝑑𝑜𝑤 𝑝𝑟𝑖𝑐𝑒 Δ𝐵 .

(previous e.g. 𝑂𝐵𝐽 7668 4.375 40 7843 .

19
INTERPRETATIONS OF SHADOW PRICE

 Shadow price focus on change in objective function value,


NOT finding the new optimal solution.

 Binding constraint: different optimal solution; resolve to


find the new optimal solution.

 Non-binding constraint: same optimal solution

 Outside the “Allowable” range, the shadow price cannot be


used, and you must re-formulate and re-solve.

20
SIMULTANEOUS CHANGES IN PARAMETERS

1. If some or all coefficients in the objective function are


changed, then you can use the 100% rule.

2. If some or all values on the RHS of constraints are changed,


then you can use the 100% rule.

3. If both coefficients and values are changed, then you have to


re-formulate and re-solve!

4. If you are not sure, then it is best to re-formulate and re-


solve!

21
100% RULE

1. A “relative change” expressed in %, which is computed as:

 In the case of an increase, the absolute value of increased


amount the associated “Allowable Increase” 100%.

 In the case of a decrease, the absolute value of decreased


amount the associated “Allowable Decrease” 100%.

2. For simultaneous coefficient changes:

 If the sum of relative changes in coefficients does not


exceed 100%, the optimal solution remains the same.

3. For simultaneous value changes:

 If the sum of relative changes in values does not exceed


100%, the current shadow prices are still valid.

4. If the sum of relative changes in either coefficients or values


EXCEEDS 100%, then you have to re-formulate and re-
solve! 22
SIMULTANEOUS CHANGES

 Q4: Suppose the profit of a standard bag decreases to $8


(from $10) and the profit of a deluxe bag increases to $10
(from $9). Does the current product production plan remain
optimal? If yes, what is the new maximum profit?

 0.7296 72.96% 100%. The current


. .
production plan remains optimal.

 New maximum profit 8 540 10 252 6840 , or


= orig profit 8 10 540 10 9 252 6840.
23
SUMMARY

Single change in
LP

Want 𝑥 0 to Change 𝑐 in Change 𝑏 in RHS


become non-zero objective function of constraint

Within Within
Use Outside Outside
allowable allowable
reduced allowable allowable
increase/ increase/
cost to find increase/ increase/
decrease: decrease:
necessary decrease: decrease:
same opt use shadow
change in 𝑐 . rewrite LP rewrite LP
soln. See price to find
See slide and re- and re-
slide 8 for new opt val.
15. solve. solve.
new opt val. See slide 18.

Binding Non-binding
constraint: constraint: same
different opt soln. opt soln.
24
SUMMARY

Multiple changes
in LP

Change 𝑐 in
Change multiple Change multiple
objective function
𝑐 in objective 𝑏 in RHS of
and 𝑏 in RHS of
function constraint
constraint

Rewrite LP
Satisfies and re-
Satisfies
100% rule: Violate Violate solve.
100% rule:
same opt 100% rule: 100% rule:
use shadow
soln and Rewrite LP Rewrite LP
price to find
find new opt and re- and re-
new opt val.
val. See solve. solve.
See slide 23.
slide 23.

At least one
All non-binding
binding
constraints: same
constraint:
opt soln.
different opt soln.
25
MEDIA SELECTION

 The Win Big Gambling Club promotes gambling junkets from a


large Midwestern city to casinos in the Bahamas. The club has
budgeted up to $8,000 per week for local advertising. The money
is to be allocated among four promotional media: TV spots,
newspaper ads, and two types of radio advertisements. Win Big’s
goal is to reach the largest possible high-potential audience through
the various media. The following table presents the number of
potential gamblers reached by making use of an advertisement in
each of the four media. It also provides the cost per advertisement
placed and the maximum number of ads that can be purchased per
week.
 Win Big’s contractual arrangements require that at least five radio
spots be placed each week. To ensure a broad-scoped promotional
campaign, management also insists that no more than $1,800 be
spent on radio advertising every week.
Medium Audience Cost per Ad Maximum Ads
Reached per ($) per Week
Ad
TV spot 5,000 800 12
(1 minute)
Daily newspaper 8,500 925 5
(full-page ad)
Radio spot 2,400 290 25 26

(30 seconds, prime time)


Radio spot 2,800 380 20
(1 minute, afternoon)
MEDIA SELECTION

 Spreadsheet formulation

27
MEDIA SELECTION

 Sensitivity Reports

28
MEDIA SELECTION

 What would be the impact if management approves


spending $200 more on radio advertising each week?

Ans:

29
MEDIA SELECTION

 Would it help Win Big if it can get out of the contractual


agreement to place at least five radio spots each week?

Ans:

30
MEDIA SELECTION

 The radio station manager agrees to run the afternoon radio spots
during some of their more popular programs. He thinks this will
increase the audience reached per ad to 3,100. Will this change the
optimal solution? Why or why not?
Ans:

31
MEDIA SELECTION

 There is some uncertainty in the audience reached per TV


spot. For what range of values for this objective function
coefficient will the current solution remain optimal?
Ans:

32
SUMMARY

 Linear Programming Formulation

 Decision variables

 Linear objective function

 Linear constraints

 Sign restrictions

 Applications

 Product mix

 Blending

 Multi-period investment

 Transportation

 Work scheduling

 Solution:

 Solver → Simplex algorithm (i.e. a computational procedure)


33

 Sensitivity Report

You might also like