American Journal of Statistics and Probability

2017; 2(1): 1-8

http://www.sciencepublishinggroup.com/j/ajsp
doi: 10.11648/j.ajsp.20170201.11

Karmarkar’s Approach for Solving Linear Programming

Problem for Profit Maximization in Production Industries:
NBC Port-Harcourt Plant
Nsikan Paul Akpan, Okeh Chibuzor Ojoh
Department of Mathematics and Statistics, Faculty of Science, University of Port Harcourt, Port Harcourt, Nigeria

Email address:
nsipaulakpan@gmail.com (N. P. Akpan), okehchibuzorojoh@gmail.com (O. C. Ojoh)

To cite this article:

Nsikan Paul Akpan, Okeh Chibuzor Ojoh. Karmarkar’s Approach for Solving Linear Programming Problem for Profit Maximization in
Production Industries: NBC Port-Harcourt Plant. American Journal of Statistics and Probability. Vol. 2, No. 1, 2017, pp. 1-8.
doi: 10.11648/j.ajsp.20170201.11

Received: December 4, 2016; Accepted: February 13, 2017; Published: April 14, 2017

Abstract: In this work, the Karmarkar’s algorithm of the interior point method is compared to the simplex method by
ascertaining the effect of interior point algorithm on linear programming problem of high number of variables and study why it
is not so popularly used in solving linear programming problems. Six (6) products of coca-cola Hellenic Port Harcourt plant
(Coke 50cl, Coke35cl, Fanta 50cl, Fanta 35cl, Sprite 35cl and Schweppes 33cl) and their raw materials (Concentrates, Sugar,
Water and Carbon (iv) oxide) respectively were studied. The data were analysed using Scilab 5.5.2 software for Karmarmar’s
approach and Tora software for Simplex method and the results are compared. The Karmarkar’s algorithm gave a maximum
profit of N70,478,116.00 giving a breakdown that the company should produce 159300 crates of Coke 35cl, 64173 crates of
Fanta 50cl, 8419 crates of Coke 50cl, 4876 crates of Fanta 35cl, 1118 crates of Sprite and 7 crates of Schweppes from the
available resources to attain the optimal solution. The Simplex gave a maximum profit of N107,666,639.51 giving only the
production of about 339482 crates of Schweppes 33cl from the available resources to attain the optimal solution.

Keywords: Linear Programming, Interior Point, Karmarkar, Simplex, Duality Gap

1. Introduction
Linear Programming according to Hoffman and Bradley requirements. Then in 1984, Karmarkar came up with
(1995) is a branch of mathematics that deals with an another interior point algorithm which he named after
important class of constrained Optimization problems in himself. This algorithm is generally accepted due to its
manufacturing industries for decision making. This is now ability to deal with huge number of variables and produce
very common in many areas of applied science. With this optimal solution. This prompt to the use of Karmarkar’s
vast application, Robere (2012) pointed out that there has algorithm in this work.
been interest in finding suitable algorithm to find the
solution to any linear programming problem. Having 1.1. Purpose of the Study
introduced linear programming by Leonid Kantorovich for To obtain the optimal solution of a linear programming
solving a production scheduling problem, the first problem using Karmarkar’s methods of an interior point
algorithm to solve this problem called the Simplex algorithm not a corner point algorithm.
Algorithm was introduced by Dantzig in 1947. This
algorithm has been working well in practice ever since it 1.2. Statement of Problem
was introduced but in the 1970s, it was found out that this
simplex algorithm work well in all possible linear programs. Linear programming models are concerned with the optimal
This lead to the introduction of the first interior point way in which available raw materials are utilized to produce
algorithm “Ellipsoid Method” by Khachiyan in 1979 the different products of a company. Whenever there is raw
though was criticised because of its inability to meet some material to be transformed into finished goods, there is need
2 Akpan Nsikan Paul and Ojoh Chibuzor Okeh: Karmarkar’s Approach for Solving Linear Programming
Problem for Profit Maximization in Production Industries: Nbc Port-harcourt Plant

to minimize the cost of production so as to increase profit on 3. Methodology

sales. The problem becomes; what quantity of each product
to be produced from the available resources (raw materials) The method employed in this work is the interior point
to minimize costs of production and maximize profit. This algorithm of Karmarkar and the Simplex method. The
work tends to develop a better optimization tool that can Simplex method is widely known so will not be talked about
tackle this problem. This will enable companies to take extensively rather the new Karmarkar method will be
advantage of this opportunity to improve in their production. discussed.
For a linear programming problem to be solved by
Karmarkar’s method manually, such problem must be in a
2. Literature Review homogenous form (Assumption 1 of karmarkar). The second
When the runtime of simplex method is compared with assumption says that the objective function at optimal has the
Karmarkar’s method, one can observe that neither of them is value zero (0).
faster than the other in all problems. The practical efficiency Consider the problem
of both Simplex and Karmarkar’s methods depend strongly Min
on the details of their implementation. The iterations
numbers required by Karmarkar’s method is higher than that Subject to =0
1 =1
of Simplex. The simplex method needs lesser iteration like
2n to 3n iterations, where n is the number of primal variables.
Thus, generally the Interior Point Methods are better for ≥0
large-scale problems Szabo and Kovacs (2003).
Karmarkar’s algorithm seems to work better on linear In the constraint, we observe that we don’t have (basic
programming problems with no initial feasible point (k=0) solution value) rather, we have 0 and the sum of all variables
and also on very large structural problems Lemire (1989). He is equal to 1.
went ahead to say that the Karmarkar’s algorithm is valid and So when the problem is not given in this form, we convert
it exhibits polynomial time complexity. Having used the it to Karmarkar’s form before we can solve it by karmarkar’s
algorithm to solve some problems, he found out that the method. Having the problem in homogeneous form, we can
algorithm produces speed ups over the simplex method. This now solve thus by the following steps:
counter-opposes the statement made by Szabo and Kovacs (a) Set k=0 {initial iteration}
that both methods have the same speed, none is faster. (b) Find = and = the product of and
( )
⇒
Further discussion on the outline of the algorithm and
(step length)
comment on the applications of this class of method by
(c) We transform the x-space problem to a problem in the
= = ( , ,…, )
Nemirovski and Todd in 2008 equally revealed that
Karmarkar’s algorithm is convenient for polynomial time and y-space. where
large constraints with variables which supports the statement = . !"#$" %&'. Then we get ( = )$"* ( ) =
made by szabo and kovacs. )$"* ( , , … , )
Vanderbei et al (1986) Concluded that Karmarkar’s (d) We get = coefficient of variables at objective
algorithm is formulated in the positive orthant instead of the function then, A = matrix of the coefficient of the
(
variables at the constraints. + = , -
simplex method. This makes it easier to conceptualize and
lead to computational simplicity. They said that this method’s 1
sliding-objective-function is replaced by a projected gradient (e) We get the Projection matrix .
. = [0 − + (++ ) +]
search for the optimum. Empirically, this leads to a decrease
in the number of iterations the algorithm requires to solve a
problem. (f) We find the projected steepest descent direction +3 (can
Karmarkar (1984) also stated in his work on “A equally be 4 ).
Polynomial-Time Algorithm for Linear Programming” that
the algorithm is not necessary to find the exact solution to all +3 = −.5
optimization problems rather it is useful to distinguish
between two types of approximate solutions. He gave an (g) The new value of y can now be obtained by =
9: ;<
instance that if X0 is the exact solution, X is the approximate 6 78 = − where ∥ +3 ∥ (norm of +3 ) =
∥ ;< ∥
solution and is the objective function also, that a strong
approximation (approximation in solution space) requires >( +3 )? + ( +3 ? )? + ⋯ + ( +3 )?
that X be close to X0 in some suitable norm. a weak (h) The Z (objective function) value will now be
approximation (approximation in objective function space) calculated as B = .
requires be close to . This weak approximation is The termination condition is as stated before: if the duality
sufficient to Karmarkar’s method and it is easier to achieve
gap is zero or negative then stop, otherwise obtain the next
numerically.
iteration until the termination condition is met.
 American Journal of Statistics and Probability 2017; 2(1): 1-8 3

3.1. The Duality Gap at Optimal 6. An artificial variable is now introduced in all the
equations, such that the sum of the coefficients in each
− C homogeneous equation will be zero and the artificial
Where C = (+ ? + ) + ? where P takes values of the
variable in the last equation is one. These artificial
( variables are to be minimized. The artificial variable to
matrix above , - and ? = [)"$*( )]? and is as stated
1 be introduced here is ?RS? S
above.
If: 4. Data Presentation and Analysis
− C = 0 or negative{strong duality}
Then The data of this work was gotten from the production and
> C{weak duality} commercial department of Coca-cola plant of Port Harcourt.
The researcher’s interest is to ascertain the products and their
3.2. Converting Linear Programming Problem to quantities to be produced in other to make an optimal profit
Karmarkar’s Form using karmarkar and comparing to simplex. The Tora
software is used to analyse the data for Simplex and the
When the problem is in linear programming form or in
scilab software is used to analyse the data for Karmarkar
standard form, we can convert it to Karmarkar’s form before
since the data cannot be analysed manually because of the
we can solve the problem manually by Karmarkar’s method.
numerous number of variables. This lead to a matrix 4x9
There are steps to be followed in order to achieve this:
which is difficult to solve manually.
Let us consider the LPP below
Max B =
Table 1. Available raw materials in stock.

GH I&5J J ≤
Raw Materials Quantity Available
(i) {primal) 1. Concentrates 4332 (units)
≥0 2. Sugar 467012 (kg)
3. Water (H2O) 1637660 (litres)
1. Firstly, we get the dual of the primal given above as: 4. Carbon (iv) oxide (CO2) 8796 (vol. per pressure)

Min B = L Source: Coca-cola Port Harcourt Plant

GH I&5J J L ≥ 5 (ii) {dual) Table 2. Available raw materials needed to produce a crate of each product.

L≥0
Carbon (iv)
Flavours Concentrates Sugar Water
oxide
Coke 50cl ( ) 0.00359 0.89 7.552 0.0135
2. We introduce the slack and surplus variables in the Coke 35cl ( ? ) 0.0042 1.12 6.539 0.0133
primal and dual problems, and then combine the Fanta 50cl ( ) 0.0021 1.044 7.671 0.007
problems. Fanta 35cl ( T ) 0.00419 0.86 6.822 0.005
3. We introduce a boundary constraint K such that any Sprite 35cl ( U ) 0.00359 0.73 6.12 0.0149
feasible solution to the equation gotten in step 2 above Schweppes 33cl ( V ) 0.00438 0.23 4.824 0.0156
will satisfy sum of all variables in the said equation less
than or equal to K and then add a slack variable to get Source: Coca-cola Port Harcourt Plant
something like: ∑ N + ∑ LN + ' = O Table 3. Available cost of production and selling price of a crate of each
4. We homogenize the equations gotten in step 3 above by
introducing a dummy variable ) (subject to the
product.

condition ) = 1 ). We then replace the equation Products

Average cost Average selling
Profit (N)
∑ N + ∑ LN + ' = O with the following equivalent
price (N) price (N)

equations: ∑ N + ∑ LN + ' − O) = 0 and ∑ N +

Coke 50cl ( ) 760.55 1050 289.45

∑ LN + ' + ) = O + 1
Coke 35cl ( ? ) 679.51 980 300.49
Fanta 50cl ( ) 762.63 1050 287.37
5. Transform the equations in step 4 above to have one Fanta 35cl ( T ) 691.91 980 288.09
variable in all the equations instead of different Sprite 35cl ( U ) 689.67 980 290.33
variables. By this, we have Schweppes 33cl ( V ) 682.85 980 317.15

N = (O + 1) N , $ = 1, 2, … , (Q + ) Source: Coca-cola Port Harcourt Plant

LN = (O + 1) RS SN , $ = 1, 2, … , (Q + ) 4.1. Karmarkar’s Approach of Solving the Problem

' = (O + 1) ?RS? S , ) = (O + 1) ?RS? S? Model Formulation:

W" B = 289.45 + 300.49 ? + 287.37 + 288.09 T + 290.33 U + 317.15 V GH I&5J J :

0.00359 + 0.0042 ? + 0.0021 + 0.00419 T + 0.00359 U + 0.00438 V ≤ 4332
4 Akpan Nsikan Paul and Ojoh Chibuzor Okeh: Karmarkar’s Approach for Solving Linear Programming
Problem for Profit Maximization in Production Industries: Nbc Port-harcourt Plant

0.89000 + 1.12000 ? + 1.04400 + 0.86000 T + 0.7300 U + 0.2300 V ≤ 467012

7.5520 + 6.5390 ? + 7.67100 + 6.82200 T + 6.1200 U + 4.82400 V ≤ 1637660
0.01350 + 0.01330 ? + 0.0070 + 0.0050 T + 0.01490 U + 0.01560 V ≤ 8796

This can be put in Karmarkar’s form by the following steps:

Step 1: Get the dual of the above primal as:
Min B = 4332L + 467012L? + 1637660L + 8796LT
'H I&5J J :
0.00359L + 0.89L? + 7.552L + 0.0135LT ≥ 289.45
0.0042L + 1.12L? + 6.539L + 0.0133LT ≥ 300.49
0.0021L + 1.044L? + 7.671L + 0.007LT ≥ 287.37
0.00419L + 0.86L? + 6.822L + 0.005LT ≥ 288.09
0.00359L + 0.73L? + 6.12L + 0.0149LT ≥ 290.33
0.00438L + 0.23L? + 4.824L + 0.0156LT ≥ 317.15
L$Jℎ: "%% !"#$" %&' &*"J$!&
Step 2: We introduce the slack and surplus variables in the primal and dual problems, and then combine the problems.
0.00359 + 0.0042 ? + 0.0021 + 0.00419 T + 0.00359 U + 0.00438 V + a = 4332
0.89000 + 1.12000 ? + 1.04400 + 0.86000 T + 0.7300 U + 0.2300 V + b = 467012
7.5520 + 6.5390 ? + 7.67100 + 6.82200 T + 6.1200 U + 4.82400 V + c = 1637660
0.01350 + 0.01330 ? + 0.0070 + 0.0050 T + 0.01490 U + 0.01560 V + = 8796
0.00359L + 0.89L? + 7.552L + 0.0135LT − LU = 289.45
0.0042L + 1.12L? + 6.539L + 0.0133LT − LV = 300.49
0.0021L + 1.044L? + 7.671L + 0.007LT − La = 287.37
0.00419L + 0.86L? + 6.822L + 0.005LT − Lb = 288.09
0.00359L + 0.73L? + 6.12L + 0.0149LT − Lc = 290.33
0.00438L + 0.23L? + 4.824L + 0.0156LT − L = 317.15
289.45 + 300.49 ? + 287.37 + 288.09 T + 290.33 U + 317.15 V
= 4332L + 467012L? + 1637660L + 8796LT
L$Jℎ: "%% !"#$" %&' &*"J$!&
Step 3: Introduce a boundary constraint K such that any feasible solution to the equation gotten in step 2 above will satisfy
sum of all variables in the said equation less than or equal to K and then add a slack variable to get something like: ∑ N +
∑ LN + ' = O

0.00359 + 0.0042 ? + 0.0021 + 0.00419 T + 0.00359 U + 0.00438 V + a = 4332

0.89000 + 1.12000 ? + 1.04400 + 0.86000 T + 0.7300 U + 0.2300 V + b = 467012
7.5520 + 6.5390 ? + 7.67100 + 6.82200 T + 6.1200 U + 4.82400 V + c = 1637660
0.01350 + 0.01330 ? + 0.0070 + 0.0050 T + 0.01490 U + 0.01560 V + = 8796
0.00359L + 0.89L? + 7.552L + 0.0135LT − LU = 289.45
0.0042L + 1.12L? + 6.539L + 0.0133LT − LV = 300.49
 American Journal of Statistics and Probability 2017; 2(1): 1-8 5

0.0021L + 1.044L? + 7.671L + 0.007LT − La = 287.37

0.00419L + 0.86L? + 6.822L + 0.005LT − Lb = 288.09
0.00359L + 0.73L? + 6.12L + 0.0149LT − Lc = 290.33
0.00438L + 0.23L? + 4.824L + 0.0156LT − L = 317.15
289.45 + 300.49 ? + 287.37 + 288.09 T + 290.33 U + 317.15 V − 4332L − 467012L? − 1637660L − 8796LT = 0

d N + d LN + G = O
Ne Ne

L$Jℎ: "%% !"#$" %&' &*"J$!&

Step 4: Homogenize the equations gotten in step 3 above by introducing a dummy variable ) (subject to the condition ) =
1). We then replace the equation ∑ N + ∑ LN + ' = O with the following equivalent equations: ∑ N + ∑ LN + ' − O) = 0 and
∑ N + ∑ LN + ' + ) = O + 1

0.00359 + 0.0042 ? + 0.0021 + 0.00419 T + 0.00359 U + 0.00438 V + a − 4332) = 0

0.89000 + 1.12000 ? + 1.04400 + 0.86000 T + 0.7300 U + 0.2300 V + b − 467012) = 0
7.5520 + 6.5390 ? + 7.67100 + 6.82200 T + 6.1200 U + 4.82400 V + c − 1637660) = 0
0.01350 + 0.01330 ? + 0.0070 + 0.0050 T + 0.01490 U + 0.01560 V + − 8796) = 0
0.00359L + 0.89L? + 7.552L + 0.0135LT − LU − 289.45) = 0
0.0042L + 1.12L? + 6.539L + 0.0133LT − LV − 300.49) = 0
0.0021L + 1.044L? + 7.671L + 0.007LT − La − 287.37) = 0
0.00419L + 0.86L? + 6.822L + 0.005LT − Lb − 288.09) = 0
0.00359L + 0.73L? + 6.12L + 0.0149LT − Lc − 290.33) = 0
0.00438L + 0.23L? + 4.824L + 0.0156LT − L − 317.15) = 0
289.45 + 300.49 ? + 287.37 + 288.09 T + 290.33 U + 317.15 V − 4332L − 467012L? − 1637660L − 8796LT = 0

d N + d LN + G − O) = 0
Ne Ne

d N + d LN + G + ) = (O + 1)
Ne Ne

L$Jℎ: "%% !"#$" %&' &*"J$!&

Step 5: Transform the equations in step 4 above to have one variable in all the equations instead of different variables. By
this, we have

N = N, $ = 1, 2, … , (Q + )
LN = RS SN , $ = 1, 2, … , (Q + )
'= ?RS? S ,) = ?RS? S?

Jℎ& ' = ? " ))= ??

Jℎ& " !& J#" ' #Q"J$ ' $&%) Jℎ& %% L$ * ' 'J&Q:
0.00359 + 0.0042 ? + 0.0021 + 0.00419 T + 0.00359 U + 0.00438 V + a − 4332 ?? =0
0.89000 + 1.12000 ? + 1.04400 + 0.86000 T + 0.7300 U + 0.2300 V + b − 467012 ?? =0
7.5520 + 6.5390 ? + 7.67100 + 6.82200 T + 6.1200 U + 4.82400 V + c − 1637660 ?? =0
6 Akpan Nsikan Paul and Ojoh Chibuzor Okeh: Karmarkar’s Approach for Solving Linear Programming
Problem for Profit Maximization in Production Industries: Nbc Port-harcourt Plant

0.01350 + 0.01330 ? + 0.0070 + 0.0050 T + 0.01490 U + 0.01560 V + − 8796 ?? =0

0.00359 + 0.89 ? + 7.552 + 0.0135 T − U − 289.45 ?? =0
0.0042 + 1.12 ? + 6.539 + 0.0133 T − V − 300.49 ?? =0
0.0021 + 1.044 ? + 7.671 + 0.007 T − a − 287.37 ?? =0
0.00419 + 0.86 ? + 6.822 + 0.005 T − b − 288.09 ?? =0
0.00359 + 0.73 ? + 6.12 + 0.0149 T − c − 290.33 ?? =0
0.00438 + 0.23 ? + 4.824 + 0.0156 T − ? − 317.15 ?? =0
289.45 + 300.49 ? + 287.37 + 288.09 T + 290.33 U + 317.15 V − 4332 − 467012 ? − 1637660 − 8796 T =0
?

d N −O ?? =0
Ne
??

d N =1
Ne

L$Jℎ: "%% !"#$" %&' &*"J$!&

Step 6: An artificial variable is now introduced in all the equations, such that the sum of the coefficients in each
homogeneous equation will be zero and the artificial variable in the last equation is one. These artificial variables are to be
minimized. The artificial variable to be introduced here is ?RS? S , we then have:
W$ $Q$f&: ?

GH I&5J J :
0.00359 + 0.0042 ? + 0.0021 + 0.00419 T + 0.00359 U + 0.00438 V + a − 4332 ?? + 4332.978 ? =0
0.89000 + 1.12000 ? + 1.04400 + 0.86000 T + 0.7300 U + 0.2300 V + b − 467012 ?? + 467008.126 ? =0
7.5520 + 6.5390 ? + 7.67100 + 6.82200 T + 6.1200 U + 4.82400 V + c − 1637660 ?? + 1637621.625 ? =0
0.01350 + 0.01330 ? + 0.0070 + 0.0050 T + 0.01490 U + 0.01560 V + − 8796 ?? + 8796.931 ? =0
0.00359 + 0.89 ? + 7.552 + 0.0135 T − U − 289.45 ?? + 281.869 ? =0
0.0042 + 1.12 ? + 6.539 + 0.0133 T − V − 300.49 ?? + 293.814 ? =0
0.0021 + 1.044 ? + 7.671 + 0.007 T − a − 287.37 ?? + 279.646 ? =0
0.00419 + 0.86 ? + 6.822 + 0.005 T − b − 288.09 ?? + 274.577 ? =0
0.00359 + 0.73 ? + 6.12 + 0.0149 T − c − 290.33 ?? + 284.475 ? =0
0.00438 + 0.23 ? + 4.824 + 0.0156 T − ? − 317.15 ?? + 313.076 ? =0
?

d N −O ?? + (O − 21) ? =0
Ne
?

d N =1
Ne

L$Jℎ: "%% !"#$" %&' &*"J$!&

This can now be solved using Scilab 5.5.2 software and then result will be obtained.
Analysis from Scilab of Karmarkar’s Method
The above data is analysed using scilab 5.5.2 for karmarkar’s method and the result is as below:
 American Journal of Statistics and Probability 2017; 2(1): 1-8 7

Table 4. Result from the analysis by Karmarkar’s Approach.

Variables Explanations Values

Iter No. of iterations 56
Exitflag 1
Fopt Objective function 70478116
Xopt No. of crates
Coke 50cl 8419
? Coke 35cl 159300
Fanta 50cl 64173
T Fanta 35cl 4876
U Sprite 35cl 1118
V Schweppes 33cl 7

The result from scilab 5.5.2

4.2. Simplex Method of Solving the Problem

Model formation
W" B = 289.45 + 300.49 ? + 287.37 + 288.09 T + 290.33 U + 317.15 V + 0G + 0G? + 0G + 0GT
GH I&5J J :
0.00359 + 0.0042 ? + 0.0021 + 0.00419 T + 0.00359 U + 0.00438 V + G + 0G? + 0G + 0GT = 4332
0.89000 + 1.12000 ? + 1.04400 + 0.86000 T + 0.7300 U + 0.2300 V + 0G + G? + 0G + 0GT = 467012
7.5520 + 6.5390 ? + 7.67100 + 6.82200 T + 6.1200 U + 4.82400 V + 0G + 0G? + G + 0GT = 1637660
0.01350 + 0.01330 ? + 0.0070 + 0.0050 T + 0.01490 U + 0.01560 V + G + 0G? + 0G + GT = 8796

This can be put in standard form as

W" B = 289.45 + 300.49 ? + 287.37 + 288.09 T + 290.33 U + 317.15 V + 0G + 0G? + 0G + 0GT
GH I&5J J :
0.00359 + 0.0042 ? + 0.0021 + 0.00419 T + 0.00359 U + 0.00438 V + G + 0G? + 0G + 0GT = 4332
0.89000 + 1.12000 ? + 1.04400 + 0.86000 T + 0.7300 U + 0.2300 V + 0G + G? + 0G + 0GT = 467012
7.5520 + 6.5390 ? + 7.67100 + 6.82200 T + 6.1200 U + 4.82400 V + 0G + 0G? + G + 0GT = 1637660
0.01350 + 0.01330 ? + 0.0070 + 0.0050 T + 0.01490 U + 0.01560 V + G + 0G? + 0G + GT = 8796
This can be solved with TORA software and results will be
obtained
5. Conclusion
Analysis from Tora of Simplex Method From the result above, it is shown that the production of
about 159300 crates of Coke 35cl, 64173 crates of Fanta
Table 5. Result from the analysis using Simplex Method.
50cl, 8419 crates of Coke 50cl, 4876 crates of Fanta 35cl,
Variables Explanations Values 1118 crates of Sprite and 7 crates of Schweppes with that
No. of iterations 2 available resources in the Port Harcourt plant of the Coca-
Objective function 107666640 cola company will make an optimal profit of N70,478,116.00
No. of crates Looking at the result, Coke 35cl is the most significant
Coke 50cl 0.0000 product that will push up profit while the production of
? Coke 35cl 0.0000 Schweppes is almost insignificant. Then the result shows that
Fanta 50cl 0.0000
the coca-cola plant in Port Harcourt should produce about
T Fanta 35cl 0.0000
339482 crates of Schweppes only with the available
U Sprite 35cl 0.0000
resources in other to make an optimal profit of
V Schweppes 33cl 339481.76
N107,666,640.00 By observation, it is noticed that the other
'ga Slack 2845.07
'gb Slack 388931.20
products were not put into consideration as to how many
'g Slack 3500.08 crates to be produced.

The result from TORA

8 Akpan Nsikan Paul and Ojoh Chibuzor Okeh: Karmarkar’s Approach for Solving Linear Programming
Problem for Profit Maximization in Production Industries: Nbc Port-harcourt Plant

Recommendations [9] Nemirovski A. S. and Todd M. J. (2008) “Interior-point

method for Optimization” Cambridge University Press,
Since Karmarkar’s approach gave a better result from the United Kingdom.
data analysed, the production team of Coca-cola Port [10] Nimrod Megiddo (1986) “Linear Programming” IBM
Harcourt is encourage to adopt this method for better Resaerch, Statistics Department, Tel Aviv University, Israel.
decision making. However, it is still uncertain as to how
problems that produce dense A matrix or dense factorizations [11] Ponnambalam et al (1989) “An Application of Karmarkar’s
Interior-point linear programming algorithm for multi-
could be solved efficiently using the Karmarkar’s algorithm, reservior operations optimization” University of Waterloo,
the direction for future research include (i) development of Ontario, Canada.
better methods to determine the initial feasible interior-point
and (ii) decomposition methods for essentially sparse [12] Richard B. and Govindasami N. (1997) “Schaum’s Outline
Operations Research” 2nd Edition. MC Graw-Hill Inc. USA.
problems but with a few dense rows/columns. If these are
achieve, solving linear problems by Karmarkar’s algorithm [13] Robere Robert (2012) “interior point methods and Linear
manually will be easier and a good alternative to Simplex Programming” University of Toronto.
method would have equally been achieved. [14] Schrijver A. (1986) “The new linear programming Method of
karmarkar” Centre for Mathematics and Computer Science,
Amsterdam, Netherlands.
References [15] Sharma J. K (2013) “Operations Research: Theory and
[1] Barnes E. R (1995) “A Variation on Karmarkar’s Algorithm Applications” 5th Edition. Mac Millan Publishers india ltd.
for Solving Linear Programming Problems” yorktown heights, India.
New York. [16] Shirish K. Shevade (2012) “Numerical Optimization” An
[2] Dantzig G. B (1947) “Linear Programming and Extension” online lecture of Indian Institute of Science. Bangalore, India.
Princeton University Press. New Jessey. [17] Szabo Zsuzsanna and Kovacs Marta (2003) “Interior-Point
[3] Gill et al (1986): On projected Newton barrier methods for Methods And Simplex Method In Linear Programming” St.
linear programming and an equivalence to Karmarkar's Univ. Ovidius constanta vol. 11 (2), 2003, 155–162. Petru
projective method, Math. Programming, 36, 183-209. Maior University of TgMures, Romania.

[4] Hoffman L. D. and Bradley G. L. (1995) “Finite Mathematics [18] Taha H. A. (2008) “Operations Research: An Introduction”
with Calculus” page (153). MC Graw Hill Inc. New York. Eight Edition. Pearson press, Dorling Kindershey (india) pvt.
Ltd. India.
[5] Jessen P. A. and Bard J. F. (2000) LP Methods: Interior point
Method. [19] Todd M. J. (1991) “The Affine-scalling Direction for Linear
Programming is a limit of projection scalling directions”
[6] Karmarkar N. (1984) “A New Polynomial-time Algorithm for Cornell University, Upson Hall Ithaca, New York.
Linear programming” AT&T Bell Laboratories murray Hill,
USA. [20] Vanderbei R. J. (1986) “Convex optimization: Interior Point
Methods and Applications.” Princeton University Press.
[7] Khachiyan, L. G. (1979): “A polynomial algorithm in linear
programming”. Sou. Math. Dokl. 20: 191-94. [21] Vanderbei, R. J, Meketon M. J, Freedman, B. A (1986) “A
Modification of Karmarkar’s Linear Programming Algorithm.
[8] Korte & Vygen on www. Wikipedia.com.