SENSITIVITY ANALYSIS

BY:-
ANUJ GANG
ANSHUL DHURANDHAR
AMAN MACHRA
ANIKET MODI
ANISH KUMAR NEMANI
ANKIT PURI
ADNAN
 What is sensitivity analysis?

Finding the optimal solution to a linear programming

model is important, but it is not the only information
available. There is a tremendous amount of sensitivity
information, or information about what happens when data
values are changed.
In order to formulate a problem as a linear program, you
had to invoke a certainty assumption: you had to know
what value the data(e.g. profit per item produced or
availability component etc.) took on, and you have made
decisions based on that data. Often this assumption is
somewhat dubious: the data might be unknown, or guessed
at, or otherwise inaccurate.
 Sensitivity Analysis
Sensitivity analysis refers to the study of the changes
in the optimal solution and optimal values of objective
function (cost/profit) due to change in the input data
coefficient (resource/labor/administration etc.).
For a discussion of sensitivity analysis, we shall
confine ourselves to following changes in the data, and
how to handle these changes.
1. Changes in the cost coefficient (cj).
2. Changes in the right hand side constraints (bi).
3. Changes in the constraint or coefficient matrix (A).
- Adding new activities or variables.
- Changing existing columns.
- Adding new constraints.
 Illustrative example
The MDG company plans production on three of their products
A, B and C. The unit profits (cost coefficients) on these products
are 2, 3 and 1, respectively, and they require two resources
(constraints) labor and material. Company`s operation research
department formulates the linear programming model for
determining the optimal solution :

Maximize : Z = 2x1 + 3x2 + x3

Subject to : 1/3x1 + 1/3x2 + 1/3x3 =< 1 (labor)
1/3x1 + 4/3x2 + 7/3x3 =< 3 (material)
Non negativity constraint : x1 ,x2 ,x3 >= 0

where x1 ,x2 ,x3 are the number of products A, B and C produced.

 Final Simplex tableau A
Here Zj = inner Cj 2 3 1 0 0 Constan
product of Cb ts
and the column Cb Basis X1 X2 X3 X4 X5
corresponding to 2 X1 1 0 -1 4 -1 1
xj in the 3 X2 0 1 2 -1 1 2
canonical
Cj Zj 0 0 -3 -5 -1 Z=8
system.
Cj Zj= relative
profit

Henceforth we`ll be conducting sensitivity analysis on the

above problem and check whether the optimal solution
and optimal values changes or not.
 Variation in the objective function
coefficients (Cj)
Variations in the coefficients of the objective functions may
occur due to a change in profit or cost of either if basic
activity or a non basic activity so we shall treat both cases
separately.
( A variable is said to be a basic variable in a given
equation if it appears with a unit coefficient in that equation
and zeros in all other equations.
Those variables which are not basic are called non basic
variables.
By applying the elementary row operations a given variable
can be made a basic variable. This is called pivot operation.
In our case, according to final simplex table x1 and x2 are
basic and the rest are non basic variables. )
Case 1 changing the objective function coefficient of a
non basic variable
In the optimal product mix, product C is not produced because of
its low profit Re 1 per unit. One may be interested in finding out
the range on the values of C3 such that the current optimal
solution remains optimum. It is clear that when C3 decreases
optimal solution remains same. Hence,
The relative profit of = C3 (c1, c2) * (Values corresponding
non basic variable x3 to column x3 in simplex table )
For table A to be optimal relative profit should be less than
or equal to zero.
Solution : Therefore, c3 (2,3)(-1,2) = c3 4 =< 0 i.e. till the
unit profit on the is less than Rs 4 it is not economical to produce
product C
 Suppose if the unit profit of the product C increases to Rs. 6 then
relative profit = 6 4 = 2, hence current product mix is not
optimal the maximum profit can be increased further by
producing product C.

Cb Basis X1 X2 X3 X4 X5 Constants
2 X1 1 0 7/2 -1/2 2
6 X3 0 1 -1/2 1
Cj Zj 0 -1 0 -4 -2 Z=10

Hence the new product mix is to produce two units of product A

and one unit of product C with maximum profit of Rs 10.
Case 2 changing the objective function coefficient of a
basic variable
It is intuitively clear that when C1 decreases below certain level ,it
may not be profitable to include product A in the optimal product
mix . Even when C1it is possible that it may change the optimal
product mix at some level. This happens because product A may
become so profitable that the optimal solution only includes
product A. hence , there is an upper and a lower limit on the
variation of C1 within which the optimal solution given by table A
is not affected.
So, To determine the range of C1 (basic variable) the relative profit
of the non-basic variables X3 , X4 and X5 will be required and as
long as these remains non-positive table A remains optimal.
 Example
The relative profit of = C3 (c1, c2) * (Values corresponding non
basic variable x3 to column x3 in simplex table )
For X3: 1-(C1, 3)(-1,2)= C1 5 <=0
For X4: 0-(C1, 3)(4,-1)= -4C1 + 3 <=0
For X5: 0-(C1, 3)(-1,1)= C1 3 <=0
As long as relative profits corresponding to new C1 remains non-
positive ,the solutions remains optimum.
The range of C1 is the intersection of the above three inequalities
,which is 3/4 <= C1 <=3 ,so table A remains optimal as long as the
variation on C1 is within the limits imposed by all non-basic variables.
Case 3 changing the price of both basic
and non basic variable
If the basic and non-basic variables changes simultaneously then to
find out whether it will change the optimal solution or not , we
have to calculate the relative profits of non basic variable
corresponding to new coefficients. As long as , these values
remains non-positive the solution remains optimal
Ex- if Z= x1 + 4x2 + 2x3
calculating relative profits
For X3 : 2-(1,4)(-1,2) = -5<0
For X4 : 0-(1,4)(4,-1) = 0
For X5 : 0-(1,4)(-1,-1) = -3<0
Hence the solution does not change.
Changing the Right Hand Side
constant(bi)
It is the study of changes in the labor or material constraint
In order to study the effect of variation in the right hand side
constants it is sufficient to verify whether the new vector of
constant in the final tableau A stays non-negative. To do this there
is no need to solve the problem again . you have seen that any
column in the final tableau (including the right hand side ) can be
obtained by multiplying the corresponding column in the initial
tableau by the inverse of basic column
The values of new right hand side constant in the final table due to
increase in B is given by
new b = B-1*(old b)
Example
B-1 = [4,-1;-1,1]
new b = [4,-1;-1,1] [2,3]
=[5,1]
Thus when the value of the constant changes from (2,3) to
(5,1) due to new constants which is a positive vector. Hence,
the optimal solution remains same. But, the optimal value of
profit changes to Rs. 13 .
Hence, it is still optimal to produce products A & B. The only
difference lies in quantity of A& B produced.
Variation in the constraint matrix (A)
The constraints or the coefficient matrix may be changed by :
a) Adding new variable or activity.
b) Changing the resources requirement of the existing
activity.
c) Adding new constraints.
Case 1:Adding new variable or activity.

Suppose the company R&D department has come out with new
product D which requires 1 unit of labor and 1 unit of material.
The new product can be sold at a profit of Rs. 3. To see the effect
of new product, we do sensitivity analysis.
Inclusion of new product is equivalent to adding a new variable say
X6 and a column (1,1) in the initial tableau. The presnt optimal
solution remains same as long as the relative profit coefficient of
the new product is non-positive. We can calculate it as
The relative profit of = C6 (c1, c2)*B-1 * (Values
non basic variable x6 corresponding to column x6 in
simplex table )
Example:
As we know that, B-1 = [4,-1;-1,1]
So , relative profit coefficient for new activity
i.e. for X6 = C6 pi * P6
Here C6 = 3, P6 = (1,1) (Given)
and pi = (5,1) (from previous slide)
C6 = 3-(5,1)*(1,1)
= -3
This indicates that producing D will not improve the value of maximum
profit. In case, it turns out that the new activity can contribute to an
increased profit(because of its C6 value being positive), then the
simplex method will be applied to determine the new optimal solution.
Case 2: Variation in the resources
requirements of the existing activities.
When the labor or the material requirements of a non-basic
activity for example: product C change its effect on the
optimal solution can be studied by following the same step as
given in case 1. On the other hand, if the constraint
coefficient of a basic activity for example: product A or
product B change , then the basis matrix itself is affected
which inturn may affect all the quantity in the final simplex
table, so it may be better to solve the linear program over
again.
 Adding a new Constraint
The addition of a new constraint to the linear
programming model leads to two cases:
Case1: The new constraint is redundant, meaning it is
satisfied by the current optimal problem e.g. Let the new
constraint be
2/3x1 + 5/3x2 + 1/3x3 <= 5
Since the optimal solution(x1=1, x2=2) satisfies the
equation, the constraint is redundant.
 Case2: The current solution violates the new constraint. In
such cases dual simplex is used to determine the optimal
solution e.g. Let the new constraint be
2/3x1 + 5/3x2 + 1/3x3 <= 3

Cj 2 3 1 0 0 Constan
ts
Cb Basis X1 X2 X3 X4 X5
2 X1 1 0 -1 4 -1 1
3 X2 0 1 2 -1 1 2
0 X6 2/3 5/3 1/3 0 0 3
 Cj 2 3 1 0 0 0 Consta
nts
Cb Basis X1 X2 X3 X4 X5 X6
2 X1 1 0 -1 4 -1 0 1
3 X2 0 1 2 -1 1 0 2
0 X6 0 0 -7/3 -1 1 1 -1

X6 row => x6 row -2/3(x1 row) -5/3(x2 row)

Since for x6 the right hand side is negative we apply dual
simplex method.
Applying dual simplex we obtain the optimal solution as
x1=10/7, x2=8/7 and x3=3/7.