13/10/2023

LINEAR PROGRAMMING

 In the previous session, we discussed:

LP Formulation
DECISION MODELLING

 LP Solution (Lingo)
MBAA TERM-II  Slack/ surplus Variable
(2023-24)  Reduced costs

 Sensitivity Analysis
SESSION 4
LINEAR PROGRAMMING  changes in the coefficients of the objective function
[SENSITIVITY ANALYSIS
AND INTERPRETATION OF SOLUTION]
 changes in the right-hand side value of a constraint

Dr. Devendra Kumar Pathak  Shadow Price/ Dual Price

(M.Tech. & Ph.D., IIT Delhi)
Assistant Professor, 2
Operations Management & Decision Sciences,
Indian Institute of Management (IIM) Kashipur

1 2

RECAP: REDUCED COST EXAMPLE 1: LINGO SOLUTION

Reduced Costs
 The reduced cost for a decision variable whose value
is 0 in the optimal solution is:
 It is the amount by which objective function
coefficient of a DV should improve (increase
for maximization problems, decrease for
minimization problems) before this DV can be a
part of optimal solution.

Max 5x1 + 7x2

s.t. x1 < 6
 The reduced cost for a decision variable whose 2x1 + 3x2 < 19
value is > 0 in the optimal solution is 0. x1 + x2 < 8
x1 > 0 and x2 > 0

3 4

3 4

1
 13/10/2023

EXAMPLE: LINGO SOLUTION EXAMPLE: LINGO SOLUTION

Max 42D1 + 87D2

s.t.
3D1 + 6D2  480 Line 1 Capacity
4D1 + 2D2  480 Line 2 Capacity 5 6

5 6

SPECIAL CASES EXAMPLE: ALTERNATIVE OPTIMAL SOLUTIONS

 Alternative Optimal Solutions

In the graphical method, if the objective function Consider the following LP problem.
line is parallel to a boundary constraint in the
direction of optimization, there are alternate Max 4x1 + 6x2
optimal solutions, with all points on this line
segment being optimal. s.t. x1 < 6
2x1 + 3x2 < 18
x1 + x2 < 7

x1 > 0 and x2 > 0

10 11

10 11

2
 13/10/2023

EXAMPLE: ALTERNATIVE OPTIMAL SOLUTIONS SPECIAL CASES

Boundary constraint 2x1 + 3x2 < 18 and objective Infeasibility

function Max 4x1 + 6x2 are parallel. All points on • No solution to the LP problem satisfies all the
line segment A – B are optimal solutions. constraints, including the non-negativity conditions.
x2
7
x1 + x2 < 7
• Graphically, this means a feasible region does not
exist.
6
A • Causes include:
5
B x1 < 6 • A formulation error has been made.
4
• Management’s expectations are too high.
3 2x1 + 3x2 < 18
• Too many restrictions have been placed on the
2
problem (i.e. the problem is over-constrained).
1
x1
1 2 3 4 5 6 7 8 9 10 12 13

12 13

EXAMPLE: INFEASIBLE PROBLEM EXAMPLE: INFEASIBLE PROBLEM

 Consider the following LP problem.  There are no points that satisfy both constraints, so
there is no feasible region (and no feasible solution).
Max 2x1 + 6x2 x2
10
s.t. 4x1 + 3x2 < 12
2x1 + x2 > 8 2x1 + x2 > 8
8

x 1, x 2 > 0 6
4x1 + 3x2 < 12
4

2
Graph?
x1
14
2 4 6 8 10 15

14 15

3
 13/10/2023

SPECIAL CASES EXAMPLE: UNBOUNDED SOLUTION

Unbounded  Consider the following LP problem.

• The solution to a maximization LP problem is
unbounded if the value of the solution may be Max 4x1 + 5x2
made indefinitely large without violating any of
the constraints. s.t. x1 + x2 > 5
3x1 + x2 > 8
• For real problems, this is the result of improper x 1, x 2 > 0
formulation. (Quite likely, a constraint has been
accidentally omitted.)

Graph?

16 17

16 17

EXAMPLE: UNBOUNDED SOLUTION

 The feasible region is unbounded and the objective
function line can be moved outward from the origin
without bound, infinitely increasing the objective
function.
x2
10

8
3x1 + x2 > 8 SENSITIVITY
6

4
ANALYSIS
x1 + x2 > 5
2

x1 18 19
2 4 6 8 10

18 19

4
 13/10/2023

INTRODUCTION TO SENSITIVITY ANALYSIS OBJECTIVE FUNCTION COEFFICIENTS

 Sensitivity analysis (or post-optimality  Let us consider how changes in the objective
analysis) is used to determine how the optimal function coefficients might affect the optimal
solution is affected by changes, within specified solution.
ranges, in:
 the objective function coefficients
 The range of optimality for each coefficient
 the right-hand side (RHS) values in constraints provides the range of values over which the
current solution will remain optimal.
 Sensitivity analysis allows a manager to ask certain
what-if questions about the problem.
 Managers should focus on those objective
coefficients that have a narrow range of optimality
Max 5x1 + 7x2 and coefficients near the endpoints of the range.
s.t. x1 < 6
2x1 + 3x2 < 19
x1 + x2 < 8 20 21

x1 > 0 and x2 > 0

20 21

EXAMPLE 1: LINGO SOLUTION EXAMPLE 1: LINGO SOLUTION

Max 5x1 + 7x2

s.t. x1 < 6
2x1 + 3x2 < 19 22 23
x1 + x2 < 8
x1 > 0 and x2 > 0

22 23

5
 13/10/2023

EXAMPLE 1 EXAMPLE 1

Max 5x1 + 7x2

s.t. x1 < 6
2x1 + 3x2 < 19
24 25
x1 + x2 < 8
x1 > 0 and x2 > 0

24 25

EXAMPLE 1 RIGHT-HAND SIDES (ROF)

Range of Optimality for c1 and c2  Let us consider how a change in the right-hand side
for a constraint might affect the feasible region
Variable Cells and perhaps cause a change in the OFV.
Model Final Reduced Objective Allowable Allowable
Variable Name Value Cost Coefficient Increase Decrease
X1 X1 5.000 0.000 5.000 2.000 0.333  Shadow Price: It is the marginal value of a resource
X2 X2 3.000 0.000 7.000 0.500 2.000
by which the OFV will improve (for a positive
Constraints shadow price) when increasing the RHS of
Constraint Final Shadow Constraint Allowable Allowable constraint by 1 unit.
Number Name Value Price R.H. Side Increase Decrease
1 1 5.000 0.000 6.000 1E+30 1.000
2 2 19.000 2.000 19.000 5.000 1.000  The range of feasibility is the range over which the
3 3 8.000 1.000 8.000 0.333 1.667 shadow price is applicable.

26 27
If OFC of X1 is increased by 1.5, how will it affect optimal
solution? Can you compute the revised OFV?

26 27

6
 13/10/2023

SHADOW PRICE EXAMPLE 1

 Graphically, a shadow price is determined by  Shadow Prices

adding +1 to the right hand side value (of Constraint 1: Since x1 < 6 is not a binding
binding constraint) in question and then resolving constraint, its shadow price is 0.
for the optimal solution in terms of the same
two binding constraints.
Constraint 2: Change the RHS value of the second
Constraint to 20 and resolve for the optimal point
 The shadow price for a non-binding
constraint is 0. determined by the last two constraints:
2x1 + 3x2 = 20 and x1 + x2 = 8.
 A negative shadow price indicates that the objective The solution is x1 = 4, x2 = 4, z = 48. Hence, the
function will not improve if the RHS is increased. shadow price = znew - zold = 48 - 46 = 2.
Max 5x1 + 7x2
s.t. x1 < 6
28 2x1 + 3x2 < 19 29
x1 + x2 < 8
x1, x2 > 0

28 29

EXAMPLE 1 EXAMPLE 1
 Shadow Prices
 Shadow Prices
Constraint 3: Change the RHS value of the third
Variable Cells
constraint to 9 and resolve for the optimal point Model Final Reduced Objective Allowable Allowable
determined by the last two constraints: 2x1 + 3x2 = Variable Name Value Cost Coefficient Increase Decrease
X1 X1 5.000 0.000 5.000 2.000 0.333
19 X2 X2 3.000 0.000 7.000 0.500 2.000
and x1 + x2 = 9.
Constraints
The solution is: x1 = 8, x2 = 1, z = 47.
Constraint Final Shadow Constraint Allowable Allowable
The shadow price is znew - zold = 47 - 46 = 1. Number Name Value Price R.H. Side Increase Decrease
1 1 5.000 0.000 6.000 1E+30 1.000
2 2 19.000 2.000 19.000 5.000 1.000
Shadow Price: It is the marginal value 3 3 8.000 1.000 8.000 0.333 1.667
Max 5x1 + 7x2 of a resource by which the OFV will
Max 5x1 + 7x2
s.t. x1 < 6 improve (for a positive shadow price)
s.t. x1 < 6
2x1 + 3x2 < 19 when increasing the RHS of constraint 30 31
2x1 + 3x2 < 19
x1 + x2 < 8 by 1 unit. x1 + x2 < 8
x1, x2 > 0 x1, x2 > 0

30 31

7
 13/10/2023

RANGE OF FEASIBILITY EXAMPLE 1

 The range of feasibility for a change in the right  Range of Feasibility

hand side value is the range of values for this
coefficient in which the original dual price Variable Cells
Model Final Reduced Objective Allowable Allowable
remains constant. Variable Name Value Cost Coefficient Increase Decrease
X1 X1 5.000 0.000 5.000 2.000 0.333
X2 X2 3.000 0.000 7.000 0.500 2.000

Constraints
Constraint Final Shadow Constraint Allowable Allowable
Number Name Value Price R.H. Side Increase Decrease
1 1 5.000 0.000 6.000 1E+30 1.000
2 2 19.000 2.000 19.000 5.000 1.000
3 3 8.000 1.000 8.000 0.333 1.667

32 33
If RHS of constraint 2 is increased by 2 units, how will it
affect OFV? Can you compute OFV from this table?

32 33

EXAMPLE

34

34