Linear Programming Problem (LPP)
Sensitivity Analysis (or post-optimality analysis)
• In an LP model, the coefficients (parameters) are assumed to be constant
and known with certainty during a planning period.
• These input parameters value may change due to dynamic nature of
the business environment.
• These parameters may raise doubt on the validity of the optimal solution of
the given LP model.
• A decision-maker, in such situations, would like to know how changes in
these parameters may affect the optimal solution and the range within
which the optimal solution will remain unchanged.
Linear Programming Problem (LPP)
Sensitivity Analysis
• Sensitivity analysis and parametric linear programming are the two techniques
that are used to evaluate the effect on an optimal solution of any LP problem due
to changes in its parameters.
– Sensitivity analysis determines the sensitivity range (both lower and upper
limit) within which the LP model parameters can vary (one at a time) without
affecting the optimality of the current solution.
– Parametric analysis is the study of measuring the effect on the optimal
solution of the LP model due to changes at a time in more than one input
parameter value outside the sensitivity range.
• It provides the sensitive ranges (both lower and upper limits) within which the LP
model parameters can vary without changing the optimality of the current
optimal solution.
Linear Programming Problem (LPP)
Sensitivity Analysis: RHS Sensitivity
• It does not begin until the optimal solution to the given LP model has been
obtained.
• The RHS sensitivity involves examining how changes in the Right hand side
coefficients of the constraints affect the optimal solution.
• Since slack variables are associated with the constraints, changes in the RHS
affect the values of the slack variables and consequently the feasibility of the
solution.
Linear Programming Problem (LPP)
Sensitivity Analysis: RHS Sensitivity
• It does not begin until the optimal solution to the given LP model has been
obtained.
Steps
• Identify the slack and surplus variables that represent each resources.
• Prepare a table extracted from final simplex table showing the quantity
column, slack variable column, positive and negative ration calculation column.
• Perform calculation:
• Lower bound = original resource value – smallest number in positive ratio column
• Upper bound = original resource value – largest number in negative ratio column
• If the positive and negative ratio column has no value then lower bound is – ∞
and upper bound is +∞
Linear Programming Problem (LPP)
Sensitivity Analysis: RHS Sensitivity
Maximize Z = 25x1+42x2+30x3
Subject to:
3x1+4x2+2x3 ≤ 60
2x1+1x2+2x3 ≤ 36
1x1+3x2+2x3 ≤ 62
X1, x2, x3, ≥ 0 Optimal Solution table
Cj 25 42 30 0 0 0
Cb Basis x1 x2 x3 s1 s2 s3 RHS (Q)
42 x2 1/3 1 0 1/3 -1/3 0 8
30 x3 5/6 0 1 -1/6 2/3 0 14
0 s3 -5/3 0 0 -2/3 -1/3 1 10
Zj 39 42 30 9 6 0
Cj - Zj -14 0 0 -9 -6 0
Linear Programming Problem (LPP)
Sensitivity Analysis: Profit Coefficients of non basic variables
Maximize Z = 25x1+42x2+30x3
Subject to:
3x1+4x2+2x3 ≤ 60
2x1+1x2+2x3 ≤ 36
1x1+3x2+2x3 ≤ 62
X1, x2, x3, ≥ 0 Optimal Solution table
Cj 25 42 30 0 0 0
Cb Basis x1 x2 x3 s1 s2 s3 RHS (Q)
42 x2 1/3 1 0 1/3 -1/3 0 8
30 x3 5/6 0 1 -1/6 2/3 0 14
0 s3 -5/3 0 0 -2/3 -1/3 1 10
Cj - Zj -14 0 0 -9 -6 0
Linear Programming Problem (LPP)
Sensitivity Analysis: Profit Coefficients of non basic variables
Consider the non-basic variable x1
Let Δ be the permissible change in the profit coefficient
Cj 25+ Δ
The current solution remains optimal as long as Cj – Zj values
Basis x1 are either 0 or negative.
x2 1/3
-14+ Δ ≤ 0
x3 5/6 Δ ≤ 14
s3 -5/3
Zj 39 25 - ∞ ≤ Cj ≤ 25 + 14
Cj - Zj -14+ Δ
Linear Programming Problem (LPP)
Sensitivity Analysis: Profit Coefficients of basic variables
Consider the basic variable x2
Its profit coefficient is 42
Let Δ be the permissible change in the profit coefficient Optimal Solution table
Cj 25 42+Δ 30 0 0 0
Cb Basis x1 x2 x3 s1 s2 s3 RHS (Q)
42+Δ x2 1/3 1 0 1/3 -1/3 0 8
30 x3 5/6 0 1 -1/6 2/3 0 14
0 s3 -5/3 0 0 -2/3 -1/3 1 10
Zj 39 42+Δ 30 9+( Δ) 34+( Δ) 0
Cj - Zj -14 0 0 -9 - ( Δ) -(34+( Δ) 0
Linear Programming Problem (LPP)
Sensitivity Analysis: Profit Coefficients of basic variables for x2 variable
Absolute value of Cj - Zj Corresponding number Positive Negative
in x2 row ratio ratio
14 1/3 42
9 1/3 27
6 -1/3 -18
Upper bound = original value – (smallest number in the positive ratio column)
= 42 – 27
= 15
Lower bound = original value – (largest number in the negative ratio column)
= 42 – (-18)
= 60
Range: 15 60
Linear Programming Problem (LPP)
Sensitivity Analysis: Profit Coefficients of basic variables
Consider the basic variable x3
Its profit coefficient is 30
Let Δ be the permissible change in the profit coefficient Optimal Solution table
Cj 25 42 30+Δ 0 0 0
Cb Basis x1 x2 x3 s1 s2 s3 RHS (Q)
42 x2 1/3 1 0 1/3 -1/3 0 8
30+Δ x3 5/6 0 1 -1/6 2/3 0 14
0 s3 -5/3 0 0 -2/3 -1/3 1 10
Zj 39 42+Δ 30+Δ -( Δ) 6-( Δ) 0
Cj - Zj -14 0 0 - ( -( Δ)) 0
Linear Programming Problem (LPP)
Sensitivity Analysis: Profit Coefficients of basic variables for x3 variable
Absolute value of Cj - Zj Corresponding number Positive Negative
in x3 row ratio ratio
14 5/6 16.8
9 -1/6 -54
6 2/3 9
Upper bound = original value – (smallest number in the positive ratio column)
= 30 – 9
= 21
Lower bound = original value – (largest number in the negative ratio column)
= 30 – (-54)
= 84
Range: 21 84
Reference Books:
• Vohra, N. D. (2006) Quantitative Techniques in Management, Third Edition. New
Delhi: Tata McGraw Hill Publishing Company Limited.
• Sharma, J. K. (2009) Operations Research – Theory and Applications, Fourth Edition.
New Delhi: MacMillan Publishers India Ltd.
• Rao, S. S. (2009) Engineering Optimization – Theory and Practice. Fourth Edition. New
Jersey: John Wiley & Sons, Inc.
• Taha, H. A. (2010) Operations Research: An Introduction, Ninth Edition. New Jersey:
Prentice Hall.