BITS Pilani presentation

BITS Pilani
Pilani Campus
 BITS Pilani
Pilani | Dubai | Goa | Hyderabad

OR Models
Sensitivity Analysis – Graphical Model
Recap
 Graphical solution – Two decision variables
 Unique solution or Infinitely many solution or No solution
 Convex space
 Feasible solution
 Extreme points
 Contour lines
 Binding & non-binding constraints
 Standard form
 Slack and surplus variables

ZG568 OR Models 3 BITS-Pilani

A Minimization Problem
Min z = x + 3y
s.t.
3x + y ≥ 15
x + 2y ≥ 12
3x + 2y ≤ 24
x, y ≥ 0

ZG568 OR Models 4 BITS-Pilani

Coordinates of the Extreme Points

y
15 Min z = x + 3y
The intersection of the 1st
& 2nd
constraints s.t.
• 3x + y = 15 & x + 2y = 12 is (x = 3.6, y = 4.2) 3x + y ≥ 15
12
x + 2y ≥ 12
3x + 2y ≤ 24
The intersection of the 1st & 3rd constraints x, y ≥ 0
• 3x + y = 15 & 3x + 2y = 24 is (x = 2, y = 9) (2, 9)

6
The Intersection of the 2nd
& 3rd
constraints
• X + 2y = 12 & 3x + 2y = 24 is (x = 6, y = 3) (3.6, 4.2) (6, 3)

x
5 8 12

ZG568 OR Models 5 BITS-Pilani

The Optimal Solution

Point Z y
Min z = x + 3y
15
s.t.
(2,9) 29
3x + y ≥ 15
(6,3) 15 12
x + 2y ≥ 12
3x + 2y ≤ 24
(3.6, 4.2) 16.2 x, y ≥ 0
(2, 9)

(3.6, 4.2) (6, 3)

x
5 8 12

ZG568 OR Models 6 BITS-Pilani

RHS of a Constraint
Consider x + 2y ≥ 12
y
15 Min z = x + 3y
What happen if RHS = 13? s.t.
• The feasible space shrinks 3x + y ≥ 15
x + 2y ≥ 12
• Optimum Objective Function value deteriorates (↑) 12
3x + 2y ≤ 24
x, y ≥ 0
What happen if RHS = 11? (2, 9)

• The feasible space enlarges 6

• Optimum Objective Function value improves (↓)
(3.6, 4.2) (6, 3)

𝜕z
Shadow price = >0 x
𝜕b
5 8 12

Improvement can be computed w/o re-solving

• Check Range of Feasibility & Shadow Price
ZG568 OR Models 7 BITS-Pilani
RHS of a Constraint
Consider 3x + 2y ≤ 24
y
15 Min z = x + 3y
What happen if RHS = 25? s.t.
3x + y ≥ 15
• The feasible space enlarges
x + 2y ≥ 12
• Optimum Objective Function value improves (↓) 12
3x + 2y ≤ 24
x, y ≥ 0
(2, 9)
What happen if RHS = 23?
• The feasible space shrinks 6
• Optimum Objective Function value deteriorates (↑)
(3.6, 4.2) (6, 3)

𝜕z
Shadow price = <0 x
𝜕b 5 8 12

Improvement can be computed w/o re-solving

• Check Range of Feasibility & Shadow Price
ZG568 OR Models 8 BITS-Pilani
RHS of a Constraint
Consider 3x + y ≥ 15
y
15 Min z = x + 3y
What happen if RHS = 14? s.t.
3x + y ≥ 15
• The feasible space enlarges
x + 2y ≥ 12
• But …. 12
3x + 2y ≤ 24
x, y ≥ 0
(2, 9)
What happen if RHS = 16?
• The feasible space shrinks 6
• But …..
(3.6, 4.2) (6, 3)

This constraint is non-binding x

• LHS = 21 > RHS = 15 5 8 12

𝜕z
Shadow price = =0
𝜕b
ZG568 OR Models 9 BITS-Pilani
Shadow Prices: Remarks
The objective function is improved by weakening the binding constraint
• Enlarging the feasible region

• LHS ≤ RHS
o Enlarging the feasible region ⇒ Increasing RHS

• LHS ≥ RHS
o Enlarging the feasible region ⇒ Decreasing RHS

≤ Constraint ≥ Constraint
Problem Type 𝜕z
Shadow price =
𝜕b
𝜕z 𝜕z
Minimization ≤0 ≥0
𝜕b 𝜕b
𝜕z 𝜕z
Maximization ≥0 ≤0
𝜕b 𝜕b
ZG568 OR Models 10 BITS-Pilani
Shadow Prices: Equality Constraint
Consider a constraint where LHS = RHS
• x + y = 100 may be considered as made up of x + y ≤ 100 & x + y ≥ 100

• At most one of them will have a nonzero Shadow Price.

• Sign of the Shadow Price can be used to infer the sign of the binding constraint

• Minimization Problem
𝜕z
o < 0 implies the Objective Function will decrease if the RHS increases
𝜕b

• Maximization Problem
𝜕z
o > 0 implies the Objective Function will increase if the RHS increases
𝜕b
ZG568 OR Models 11 BITS-Pilani
Objective Function Coefficients

y
15 Min z = x + 3y
s.t.
3x + y ≥ 15
12
x + 2y ≥ 12
3x + 2y ≤ 24
x, y ≥ 0
(2, 9)

What happens if the Black Line rotates? 6

• Range of Optimality (3.6, 4.2) (6, 3)

x
5 8 12

ZG568 OR Models 12 BITS-Pilani

Reduced Cost
The reduced cost for a decision variable whose value is 0 in the optimal solution is:
the amount the variable's objective function coefficient would have to improve
(increase for maximization problems, decrease for minimization problems) before this
variable could assume a positive value.

• The reduced cost for a decision variable whose value is ≠ 0 in the optimal solution
is 0.

ZG568 OR Models 13 BITS-Pilani

Takeaway

 Sensitivity Analysis
 Changing objective coefficient in 1 decision variable
 Range in optimality
 Reduced cost
 Change in RHS of one constraint
 Range of feasibility
 Shadow price (Dual price)
 Excel Solver Sensitivity Report

ZG568 OR Models 14 BITS-Pilani

 BITS Pilani
Pilani | Dubai | Goa | Hyderabad

Thank you!!