See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/330775044

Simplex method to Optimize Mathematical manipulation

Article · February 2019

CITATIONS READS

2 10,679

2 authors:

Mohammad Rashid Hussain Mohammad Equebal Hussain

King Khalid University Suresh Gyan Vihar Univeristy, Jaipur
51 PUBLICATIONS 449 CITATIONS 14 PUBLICATIONS 19 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Mohammad Rashid Hussain on 01 February 2019.

The user has requested enhancement of the downloaded file.

 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-5, January 2019

Effect of Seven Steps Approach on Simplex

Method to Optimize the Mathematical
Manipulation
Mohammad Rashid Hussain, Mohammed Qayyum, Mohammad Equebal Hussain

 linear constraints of linear program. The slack variables

Abstract: the Simplex method is the most popular and always have a +1 coefficient when the model is in standard
successful method for solving linear programs. The objective form. For optimality the slack variables are introduced. By
function of linear programming problem (LPP) involves in the using the slack variables concept, the coefficient of objective
maximization and minimization problem with the set of linear function can be modified. According to the nature of problem
equalities and inequalities constraints. There are different it can be easily implemented on computer and these methods
methods to solve LPP, such as simplex, dual-simplex, Big-M and
overcome the limitations of graphical method and the
two phase method. In this paper, an approach is presented to
unnecessary iterations in the search can be skipped. We
solve LPP with new seven steps process by choosing “key element
rule” which is still widely used and remains important in proposed an approach of seven steps process which resolved
practice. We propose a new technique i.e. seven step process in certain complication encountered in the application of the
LPP for the simplex, dual-simplex, Big-M and two phase simplex method and feasible solution is obtained in iteration
methods to get the solution with complexity reduction. The first and optimal solution can be obtained. For the efficient
complexity reduction is done by eliminating the number of solution of large scale LPP some techniques and
elementary row transformation operation in simplex tableau of computational results have been introduced to predict
identity matrix. By the proposed technique elementary permanent basic and non-random basic variables to
transformation of operation has completely avoided and we can implement in linear programming codes to save the
achieve the results in considerable duration. computation time by eliminating the number of elementary
Index Terms: Linear programming problem (LPP), Key row transformation operation. In implementation part(V),
element (KE), Key column (KC), Key row (KR), Profit per unit our proposed approach have introduced in detail with our
(PPU), Random variables (RV), Linear Gaussian Random seven steps process, by using last two steps in simplex
variables (LGRV), standard deviation (SD), Probability Density tableau, the elementary row transformation operation are
Function (PDF) completely eliminated.
The general form of a linear program is Max (Z) =C TX
Subject to AX ≤b, X≥0
I. INTRODUCTION Where A is a mxn matrix, C is a n-vector and b is an
To solve a LPP, simplex method is the popular and widely m-vector.
used method. Simplex model in Simulink for ease in (A)mxn Xi≤bmx1, Xi; i=1,2,….,n.
visualization and simulation in System Generator are used to
achieve a fast implementation. It is efficient tableau based
representation and the clock frequency achieved by such
design is compared with that in general purpose software We get linear system with 5 variables and 3 equations.
[26]. There are some certain steps are needed to solve LPP Initial simplex tableau:
using Simplex method to implement in Standard form and it
is necessary to linear programs before solving the optimal X1 X2 S1 S2 P
solution. There are some important requirements for the
solving of LPP and to optimize: (1) If the objective function is S1 2 2(KE) 1 1 0 0 8
in minization form, it has to change to maximization (2) All KR
linear constraints must be ≤ inequality, (3) All variables must 2
be non-negative. The above three requirements must be S2 3 0 1 0 12
-3
satisfied by converting given linear program by using basic
algebra and substitution then determine slack variables i.e. P -1 0 0 1
To change inequality constraints to equality constraints there
are some additional variables have been introduced into the KC

Revised Version Manuscript Received on 28 December, 2018.

Mohammad Rashid Hussain, Department of Information Systems,
College Of Computer Science, King Khalid University, Abha, KSA
Mohammed Qayyum, Department of Computer Engineering, College Of
Computer Science, King Khalid University, Abha, KSA
Mohammad Equebal Hussain, Department of Computer Science, Suresh
Gyan Vihar University, Jaipur, India.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E1948017519/19©BEIESP 34 & Sciences Publication
 Effect of Seven Steps Approach on Simplex Method to Optimize the Mathematical Manipulation

X1 X2 S1 S2 P variables(RV) and constraints will be more then manually

X1 1 1/2 1/2 0 0 4 cannot be solve, so we need the help of computer to solve it. It
S2 2 3 0 1 0 12 is difficult to determine the value of M that has been resolved
through two phase method.
P -3 -1 0 0 1 0
III. LITERATURE REVIEW
In literature, Benichou, M., J. M. Gauthier, G. Hentges, G.
X1 X2 S1 S2 P Ribi`ere [7] and Hoffman, A., A. Mannos, D. Sokolowsky, D.
Wiegmann [11] have introduced the algorithmic techniques
X1 1 1/2 1/2 0 0 4 and computational experiences to the efficient solution of
S2 0 2 -1 1 0 4 large scale LPP and Van Roy, T. J., L. A., Wolsey[13] shows
P 0 1/2 3/2 0 1 12 automatic reformulation techniques to solve mix LPP and
how this reformulation techniques are important for
There is no negative indicator in P , so stop..!! manufacturing firm by applying the profit Preference
X1=4, X2=0 & Max (Z) =12 Scheduling in LPP’s have been implemented by Charnes, A.,
In simplex tableau, the column of their initial basic Cooper, W.W.[20], the proposed seven steps algorithm
variables, should see permutation of column of identity solves constraints integer programs introduced by
matrix, for which converting the key element to unity by Achterberg, T[8] and the process to find key element(KE), it
dividing its row by the leading element itself and all other is the KE selection methods of LPP introduced by Harris, P. J.
elements in the column to zero. J[10] and to optimize the linear programming problem by
In our proposed approach, seven steps process of last two using polynomial-time algorithm to reduce the complexity by
steps, row entry in next simplex tableau with respect to Dimitris Bertsimas and John Tsitsiklis[12,15] , An
replaced random variables and enters other than key row. Alternative Method for LPP, Beale, E.M.L [19] and have
The elementary row transformation operation is completely been consider a mathematical computational techniques to
eliminated, which have been proved in “part V implement over LPP to reduce the polynomial time
“implementation. So, the proposed approach will reduce the complexity and improves functionality of LPP’s.
complexity of maintaining unit matrix by using elementary Spielman and Teng’s JACM Paper “Smoothed analysis of
row transformation process. algorithm: why the simplex algorithm usually takes
polynomial time” [32]. The application of smoothed analysis
II. BACKGROUND algorithm is a simplex method, which found the complexity
with certain condition of standard deviation ( ) of Max ZTX,
Solutions of LPP, there are many types of Simplex
Subject to; (A+ G)x≤y, where ->0 for worst case
methods that have been developed which offer some
complexity and is ao large that of swamps out A. Based on
improvement such as reduction in the number of simplex
inputs and its domains, authors have found C-complexity
iterations and the number of computations in each iteration.
worst case, average case and smoothed C-complexity have
There are several techniques have been proposed. For
found through the concept of Gaussian random variable
example, Maximizing of a Linear Function of Variables
(GRV) on which mean=0 and variance=1. Linear
Subject to Linear Inequalities and Computational Algorithm
combination of linear GRV (LGRV’s) of X1, X2,…, XL IS
of the Simplex Method are proposed by Dantzig, G.B [21,
X=a1X1+ a1X1+….+ aLXL. So, mean of GRV (µ) = E(X) =
22], as well as by Dr. R.G. Kedia [5] and Lemke, C. E [6]
given a concept of A New Variant of Simplex Method and and E {(X- µ)2} = 2, , where the
The dual simplex method of solving the LPP through new range of standard deviation(SD), have considered
introduced search technique. Karmarkar [30] developed a , one criticsm of smoothed complexity, under the
polynomial projection approach that is developed version of relative permutation, an input is mapped to constant multiple
the ellipsoid algorithm [29], first established LPP can be of itself,
solved in polynomial time but it performs poorly in practice.
However, the simplex method is the popular to solve LPP So, the linear combination of GRV have an important
models and it is very efficient in practice, Borgwardt [31] properties in the content of communication system. When
proved that its expected number of iterations is polynomial noise is modeled as GRV which is a function of time and in
when it is applied for practical problems. In optimization the context of wireless communication channel is modeled as
problems, the Linear programming concept is applied to a a complex GRV. GRV property have used to obtain the worst
large variety of scientific and industrial computing case and average case complexity of simplex algorithm on
applications, there are several method have been introduced running time with input of form Max ZTX, subject to
to solve LPP, out of which some methods have disadvantages . To obtain the worst case analysis, simplex
and to sort out these issues new method have been algorithm, and to obtain average case complexity
introduced, the seven steps simplex algorithm is applicable to be so large that Swamps out A, So a smoothed analysis of
all the introduce Simplex method, dual simplex method, algorithm is based on Probability Density Function(PDF) to
Big-M method, Two phase method and for some of the other reduce the complexity of LPP.
method also, Big-M method have some disadvantages,
Whenever we compare M with any other numeric value, M
always will be greater, when number of random

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E1948017519/19©BEIESP 35 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-5, January 2019

In our proposed approach the concept of LPP have Phase one of the simplex method handles the computation of
reduced by removing of elementary row transformation and an initial feasible basis, which is handed over to phase two,
maintaining unit matrix in simplex tableau. the simplex method as we described it so far, T. Kitahara
There are two artificial variable technique Big M and Two and S. Mizuno [1,2] have introduced a bound for different
Phase method that we use to find the starting basic feasible basic and basic feasible solution which have been generated
solution and solve the LPP by using simplex method. In Big by simplex method with selection rule of incoming variables
M method, first we need to write the standard form of with KE closes an incoming variable whose reduced cost is
objective function and its constraints by adding some slack negative at each iteration and also introduced a bound for the
and artificial variables to get the initial basis and accordingly dual simplex method for LPP having optimal solution that
write the decision variable, under xj we consider all the have been shown some basic result when it is applied to
variables, like decision, slack or surplus and artificial specified LPP. T. Kitahara and S. Mizuno, Divya K. Nadar
variables to get the initial basic feasible solution, we required and V. Klee and G. J. Minty [3, 4, and 17] have introduced
to make all decision variable to zero and get the result of some application, properties and evaluation of computational
introduced variable. There is one disadvantage exist in amount of the simplex method and shown how good is the
Big-M method, i.e. whenever we compare M with any other simplex algorithm? In this paper, we addressed seven steps
numeric value, M will always be greater, M is very large simplex algorithm to solve LPP to reduce complexity over
value i.e. near to infinity. For big M, to get the optimal mathematical computation.
solution it contains any artificial variables in the positive Two phase method:
values if and only if the problem is not feasible, so artificial In the initial phase, we are not taking the actual
variable would not be a part of any feasible solution. coefficients of objective function, we are creating one
When number of RV and constraints will be more, than it auxiliary objective function, whatever slack and surplus
is very difficult to solve manually to compute the value of M, variables and the original decision variables takes.
we need the help of computer to solve it, another method i.e. Coefficient of objective function including slack and surplus
two phase method have been introduced to sort out this zero, and coefficient of artificial values, we add coefficient as
issues. -1.
In phase-I, the aim to eliminate artificial variables from the
A. Restriction to Normalize
basis and calculate initial basic feasible solution.
Slack, surplus and artificial variable are added to change the In phase-II, we apply simplex method with original
inequalities to equalities for equations. coefficient of objective function and not imposing penalty
Table 1 Normalize restrictions with Big-M method.
Inequality type Variable that appears Two phase implementation
≥ - Surplus + Artificial i. Convert each of the constraints into equality
= + Artificial constraints.
≤ + Slack ii. New auxiliary objective function.
The following steps of algorithm are:
i. Right hand side should be positive by multiplying (-) added with all artificial variables.
inequality constraints. Max Z*=0 Zero assigned for all artificial variables
ii. In case of minimization, multiply the objective by -1 to Max Z*< 0 Positive assigned at least 1 artificial variables
convert into maximization. Apply the two phase Simplex algorithm.
iii. For > constraints, use surplus and artificial variables. Suppose , the phase-I ends.
iv. Select value M and term the objective form as -M with
artificial variables. a) Max Z*=, All the artificial variables disappears from the
v. In ≤ constraints, slack variables are added to equalize it. basis and we will obtain basic feasible solution.
b) Max Z*= 0, one or more artificial variable appear in the
B. Background study of Big-M and Two-Phase method basis with zero value, will obtain basic feasible solution.
c) Max Z *< 0, one or more artificial variables appears in
Subject to. basis with positive value, not obtain any basic feasible
solution for the problem.
At the end of phase-I, if case (a) or (b) occur, then it will
To get the initial basis, we required to add artificial move for phase-II else end of the phase-I.
variable x5, first to write in standard form. Phase-II: There is no basic feasible solution for (c)
Assign actual coefficient of the variables of objective
Subject to. function, we are taking original objective function.
Max Z=C1x1+C2x2+C3x3+……
Now apply simplex algorithm to get the solution.
Initial basic feasible solution x4=5, x5=12.
After applying Big M method through proposed algorithm,
the solution will be x1=3, x2=2, x3=0. Maximize Z= 8
Due to disadvantages in Big M method, another method
has been introduced. i.e. two phase method.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E1948017519/19©BEIESP 36 & Sciences Publication
 Effect of Seven Steps Approach on Simplex Method to Optimize the Mathematical Manipulation

Examples for two phase method: Conditions to reach up to the optimality; for maximization
Example 1, Solution exist problem; all Cj-Zj ≤0; for minimization problem; all Cj-Zj ≥
Figure 1 shows our model of Simplex Solver Conditions to
Subject to. reach up to the optimality for maximization problem all C j-Zj
≤0 and for minimization problem all Cj-Zj ≥0. Simplex
To get the initial basis, we required adding artificial variable, iteratively searches for the optimal solution and checks the
x5 is only the artificial variable. feasible region for its computation. At the end of the first
iteration act as starting point in the next step of stopping
Subject to. condition, i.e. the condition to reach up to the optimality.
From step 1 to step 4, we find out the key column, i.e. the
highest positive value of Cj-Zj, key row (KR), i.e. the lowest
After applying simplex algorithm over phase-I and phase-II. positive value of ratio solution to the element of key column,
x1=3, x2=2, x3=0 and Zmax=8 for the first iteration, finally we get a KE, i.e. the intersection
Example 2. No solution exist point of key column and KR, then we move to the next
simplex table to reach up to the condition of optimality. In
Subject to. Step 5. Replace a random variable of KR with the variable of
key column and update its respective Cb with the coefficient
of objective function of key column and then we move for
Condition with constraints: if the constraints form ≤, Either next step i.e. Step 6. For row entry in next simplex table with
of these two method use to solve LPP, Two phase or Big-M respect to replaced random variable(key rows entry in
method and required to add artificial variables. previous simplex table /KE of previous simplex table) and the
last step to reach up to the optimality condition is Step 7, i.e.
the New entries in other than KR ( Previous values in other
Subject to. than key row-{New entries in key row*element of key
column (other than element of KR} of its respective value), at
the end of the algorithm we check the stopping condition,
Phase-I: Apply simplex algorithm For Maximization Cj-Zj ≤0 and for Minimization Cj-Zj ≥0.
When, Cj-Zj ≤0 stop iteration of phase-I, the basis contains So, the KE is normalized and other values of key column are
one artificial variable x5=4, positive value. cancelled.
If the artificial variable present in the basis and its value in
A. Duality theory/ Dual simplex method to reduce the
the basis is positive, then no solution exist.
solution complexity
So, No solution exist.
Example 3. Unbounded solution exists. To find out the solution of LPP thorough the method of
duality, first have to check either the given problem is in
Subject to. canonical form or not, if it is not in canonical, change in
canonical the apply the Simplex algorithm to solve it.
Initial basic feasible solution, where the feasibility condition
Two artificial variables required to add with constraints. is always satisfied whenever forming initial simplex table.
Here, we check either optimality condition is satisfied or not,
Subject to. if optimality condition is not satisfied then we change the
basis and going to next iteration, here in each iteration
feasibility condition is maintained. Dual Simplex method is
When Cj-Zj ≤0 for all j, Phase-I ends here. just the opposite of simplex methods, i.e. we are starting with
Note: in phase-II, the ratios are becomes negative. Both the initial optimal condition, i.e. optimality satisfied but
ratios are negative. So, unbounded solution exists. feasibility may not be satisfied, so, in each iteration , we are
changing to basis and trying to check either feasibility
IV. SYSTEM DESIGN condition is satisfied or not. In each iteration optimality must
be satisfied. If any LPP which has n variables and m
constraints, then in dual we have just opposite on, m
variables and n constraints. In dual problem, we have to
minimize the consumption of the resources in subject to the
condition and subject to the profit maximization constraints.
Minimize the consumption of the resources subject to the
maximization of the constraints.

Table 1 Framework for Overall Assessment methodology

based on seven steps Algorithm

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E1948017519/19©BEIESP 37 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-5, January 2019

We use dual method to reduce the solution complexity. D. If the problem is already in canonical form
Features: Primal
i. Feasible solution of dual model provides a bound on the Maximize
objective of original primal problem. Subject to.
ii. Optimal solution of a dual is equal to optimal solution of
the primal problem.
iii. Dual of dual model becomes original primal problem.
B. Primal and its dual Problem is in the form of Maximization.
So, it is already in canonical form.
1. Primal and it’s dual with ≥ sign. Here no. of constraints=3, No. of variables=2
Dual
Primal Let us assume dual variables as v1, v2 and v3.
Minimize
Subject to. Subject to.

Dual
Here no. of constraints=2 No. of variables=3
Subject to.
E. If the problem is not in canonical form
Primal
2. Primal and it’s dual with sign.
Primal Subject to.

Subject to.

Here no. of variables=5

Dual No. of constraints=2
Here, all the constraints are in equality form. So, first it’s
Subject to. required to change in canonical form to find its Dual.

C. General structure to convert from primal to dual Subject to.

Primal
Max
Subject to.

Let us assume dual variables as w1, w2, w3 and w4.

Dual
Dual
Min
In the given objective function, x1 and x2 are not present.
Subject to.
So, the first two constraints of dual will be ≥ 0.
Subject to.

Convert:
Min

But as per the condition of duality, the no. of variables and

constraints must be 2 and 5.
If So, our assumptions are w1-w2=v1 and w3-w4=v2, here v1 and
v2 are two positive numbers.
Results in different forms of LPP
So, v1 and v2 are unrestricted in sign.
Symmetric LPP ‘≤’ or ‘≥’
Un-symmetric LPP ‘=’
Subject to.
Mixed LPP ‘≤’ or ‘≥’ or ‘=’
Theorem1: if any constraint in primal is strict equality then
corresponding dual variable is unrestricted in sign. a≥0, b≥0,
C=a-b, here a, b and c are positive variables, then c either 0 or
positive or negative. So, c is unrestricted in sign.
Theorem 2: if any variable of the primal model is unrestricted
in sign, then the corresponding constraints of the dual will be
equality.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E1948017519/19©BEIESP 38 & Sciences Publication
 Effect of Seven Steps Approach on Simplex Method to Optimize the Mathematical Manipulation

F. Canonical problem with unrestricted in sign VI. STEPS OVERALL SIMPLEX METHOD
Primal 1. If the objective functions of given problem in
minimization form, convert it to maximization problem.
Subject to. 2. Introduce the slack variables to form the basis vectors and
construct the usual simplex table.
3. XBi=> initial basic solution.
4. Compute Zj-Cj:
(i) If Zj-Cj ≥0 for all j (optimality condition) and XBi >0 for all
Let us assume
I (feasibility condition) both exist, then the corresponding
So,
solution is the optimal basic feasible solution.
(ii) If at least one Zj-Cj <0 => the dual simplex method is not
Subject to.
applicable.
(iii) If Zj-Cj ≥0 for all j and at least one random variable
XBi<0 exist, go to step 5.
5. Dual simplex method is the mirror image of the simplex
Now it is in canonical form. method, at first we have to check which one is the
Dual departing vector, vector which has to remove from the
Let us assume dual variables as v1’, v1”, v2 and v3. basis.
So, its corresponding dual is Select the most negative value of XBi.XBr= Min { XBi,
XBi<0}, ar to the departing vector.
Subject to. 6. Check yrj (row value) for all j.
(i) If yrj>0 for all j => No feasible solution exist.
(ii) If yrj<0 for at least one j.
(Zk-Ck)/ yrk= Max {(Zj-Cj/ yrj; yrj<0}, corresponding vector ak
inter into the basis.
Let us assume
VII. ARTIFICIAL CONSTRAINTS METHOD FOR
Subject to INITIAL BASIC FEASIBLE SOLUTION
If , introduce new constraints
and sufficiently have large
=> has maximum
value.
V. RESULT BASED ON DIFFERENT ),
ASSUMPTIONS Substitute this value in the objective function and
constraints to get modified one and it will ensure that the
Primal Problem Dual Problem Conclusion
Finite optimal solution for
optimality condition satisfied.
Feasible solution Feasible solution
both primal and dual.
No Feasible Primal objective function is A. When is satisfying the condition
Feasible solution
solution unbounded
No Feasible Dual objective function is
Feasible solution Subject to.
solution unbounded
No Feasible No Feasible
Solution does not exist
solution solution

Table 2 Result based on different primal-dual Canonical form of given LPP.

assumptions
Subject to.
Dual simplex method is the mirror image of the simplex
method, like dual method is the mirror image of the primal
method. If for all j but (i.e negative) and
For corresponding vector ‘b’, AX=b, in simplex, b is always (i.e. negative), so, optimal solution is
positive but in dual simplex b must be negative. infeasible.
In dual simplex method we don’t use any artificial variable. For leaving into the basis: Min . So, x4
So, computational process and effort get reduced. In dual will leave the basis.
simplex method, initially it is not required to check feasibility For entering into the basis: Max
condition but it is important to check optimality condition is which corresponds to the
satisfied or not, optimality condition must be satisfied. vector x2, therefor x2 will enter into the basis.
AX=b, Zj-Cj≥0 for all j, aj & cj does not depends on vector b.
In initial table, optimality condition must be satisfied, but the
feasibility condition may not be satisfied, if optimality
condition does not exist, we cannot apply dual simplex
method.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E1948017519/19©BEIESP 39 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-5, January 2019

So, KE (intersection point of the departing and entering Subject to.

vectors) is -7, again , but feasibility condition
not satisfied, because one of the XBi value is negative, here,
Here,
departing vector x3 and entering vector x1. Standard form:
Again in next Simplex table.
For all j and XBi≥0 for all i,
The result of xi*, x2* and z* is to be calculated.
Subject to.

B. When is not satisfying the condition

Subject to. Here,

Corresponding canonical form by introducing slack Subject to.

variables.

Subject to.

For all i. Cj= coefficient of objective function of standard

form.(After adding the slack variables with zero
C. If for all j
coefficient)
In initial table, optimality condition is not exist.
Number of constraints=number of slack variables.
So, add artificial variable.
If Add artificial variable xM, it becomes
No. of decision variables in constraints=no. of decision
variables in objective functions.
Max This corresponds to k=no. of constraints= no. of slack variables.
variable x1, so replace x1 by M- x2+xM. Table 1. 1st iteration
Cj (Coefficient of C1 C2 … Cj 0 … 0
Subject to. Objective function) => .
BV PPU SV Pj P1 P2 … Pj s1 … sk Rk
(Cb) (P0 = => = = =
mk) x1 x2 xj
Formulate initial basic simplex table and check in modified s1 0 m1 a1 a2 … … 1 0 0 R1
form either optimality condition exist or not. s2 0 m2 b1 b2 … … 0 1 0 R2
s3 0 m3 c1 c2 … … 0 0 1 R3
In corresponding table.
… … …. … … … … … … … …
Max {-,-} =? x2 is entering vector, so x2 is entering and x3 is .
departing from the basis. sk 0 mk … … … … … … … Rk
X4 vector is departing & xm vector is entering into the basis. Cj-Zj C1 C2 … Cj 0 … 0
and for all j. Table 3 Formal structure of simplex table (1st iteration)
So, here satisfying both optimality as well as feasibility
condition. Stop here and delete the row corresponding to the In table 3, bolded PPU, BV and SV in top row of the
artificial variable. tableau states Profit per unit (PPU), RV and solution value of
The result of xi*, x2* and z* is to be calculated. the constraints, Pj and Cj represents the objective function
If optimality condition not satisfied initially, in this case also variables of standard form and its coefficients.
using the artificial constraints. The rest of rows represent linear constraint variable
We are reconstructing and reformulating the problem by coefficients from LPP.
introducing the artificial constraints and slack variables to Step 1: for simplex table: 1. Zj=0, for all j
make the artificial constraints as equality constraints and Key column= select a column with maximum coefficient
replacing one variable from here which satisfies the value of objective function= column of (Maxm(Cj)).
maximum of criteria and reformulating the For next simplex table.
problem, and in reformulated problem, once we are Before finding key column, we required to find Zj.
constructing the initial table to find the optimality condition
is satisfied.
n=no. of variable of standard objective function.
VIII. IMPLEMENTATION Key column= highest positive
value of Cj-Zj.
Step 2: Ratio (Rk) = right hand
side value of given

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E1948017519/19©BEIESP 40 & Sciences Publication
 Effect of Seven Steps Approach on Simplex Method to Optimize the Mathematical Manipulation

constraint/coefficient of same constraints variable (variable The problem is of maximization. So, we required to add slack
of key column) =SV/Element of key column= mk/ Element of variables in both the inequality equations.
key column.
Rk=1= mk/coefficient of respective variable which have been
considered for step 1(its respective key column element) Subject to. Subject to. Subject to.

.
. In given question, the number of coefficient of objective
. function is equal to the number of decision variables in
Step 3. Key row (KR) = Select a lowest positive value of Rk constraints.
(k=1, 2,…) to find a KR. i.e. lowest positive value of Rk. Table 1. 1st iteration
Cj (Coefficient of Objective 3000 2000 0 0
Step 4. Key element (KE) =Element of key column function) =>
Element of key row. BV PPU( SV Pj P1 P2 P3 P4 Rk
=Element of column (maxm Cb) (P0=mk) => = x1 = x2 = =
(Cj-Zj)) Element of row (lowest (Rk)) s1 s2
P3= 0 m1=180 5 2 1 0 180/
Now, move to next simplex table for further process. s1 5=36
Step 5. Replace a random variable of KR with the variable of P4= 0 m2=135 3 3 0 1 135/
key column and update its respective Cb with the coefficient s2 3=45
Cj-Zj 3000 2000 0 0
of objective function of key column.
Step 6. Row entry in next simplex table with respect to Table 5 System design, 1st iteration
replaced random variable=key rows entry in previous Step1. Key column (KC) =Highest positive value of
simplex table /KE of previous simplex table. Cj-Zj=3000
Step 7: New entries in other than key row= Previous values in Step2. Ratio=S.V/Element of key column= (180/5,135/3) =
other than key row-(New entries in key row*element of key (36, 45)
column (other than element of KR) of its respective value. Step3. Key row (KR) =lowest positive value of Ratio=36
End of algorithm (i.e. stopping condition):
Step4. Key element (KE) = Element of key column 
Maximization and Minimization
Element of key row.
So, the KE is normalized while the other values of the key = Element of column (maxm (Cj-Zj)) 
column are cancelled. Element of row (lowest (Rk)) = 5
Examples- For the special case of two variables and three Now, move to next simplex table for further process.
constraints, it can be explicitly written as:
Simplex Table 2.
Subject to. Cj (Coefficient of 30
2000 0 0 Rk
Subject to. Objective function) => 00
SV P1 P4
PPU Pj P2= P3=
BV (P0 = =
(Cb) => x2 s1
=mk) x1 s2
5/ 0/ 36/(2
P1= m1 =
A new introduced seven steps simplex method concept have 3000 5= 2/5 1/5 5 /5)=
x1 36
resolved certain complication encountered in the application 1 =0 90
of the simplex method and optimal solution can be obtained 27/(9
P4= m2 =
after the feasible solution which is in first iteration. 0 0 9/5 -3/5 1 /5)=
s2 27
Table 1. 1st iteration 15
Cj (Coefficient of 3 2 0 0 0 Cj-Zj 0 800 -600 0
Rk
Objective function) =>
BV PP SV Pj P1 P2 P3 P4= P5= Table 6System design, stopping condition
U (P0= = = = = s2 s3 Step5. Replace a random variable of KR with the variable of
(Cb) mk) > x1 x2 s1 key column and update its respective Cb with the coefficient
P3= 0 m1= 2 1 1 0 0 18/2 of objective function of key column.
s1 18 =9 Cb=3000 with variable x1
P4= 0 m2= 2 3 0 1 0 42/2 Step6. Row entry in next simplex table with respect to
s2 42 =21 replaced random variable=key rows entry in previous
P5= 0 m3= 3 1 0 0 1 24/3 simplex table /KE of previous simplex table.
s3 24 =8 New entries in key row = key row in old
Cj-Zj 3 2 0 0 0 table/KE=(180/5,5/5,2/5,1/5,0/5)=(36,1,2/5,1/5,0)
Table 4Example of 1st iteration based on table 3
After 1st iteration, the next process is to check the optimal
solution of maximization LPP model.
Examples-

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E1948017519/19©BEIESP 41 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-5, January 2019

Step7: New entries in other than key row= Previous values in 3. T. Kitahara and S. Mizuno: New Evaluation of Computational Amount of
the Simplex Method (Japanese), Technical Report No. 2011-8, August,
other than key row-(New entries in key row*element of key 2011.
column (other than element of key row) of its respective 4. Divya K.Nadar : Some Applications of Simplex Method, International
value={135-(36*3), 3-(1*3) ,3-(2/5*3), 0-(1/5*3), Journal of Engineering Research and Reviews ISSN 2348-697X (Online)
Vol. 4, Issue 1, pp: (60-63), Month: January - March 2016, Available at:
1-(0*3)}={27,0,9/5,-3/5,1}. www.researchpublish.com
To create an initial table, steps 1 to 4 are required to 5. Dr. R.G. Kedia: A New Variant of Simplex Method, Volume-3, Issue-6,
implement. December-2013, ISSN No.: 2250-0758 International Journal of
Engineering and Management Research Available at: www.ijemr.net
To create table two and it’s onwards, steps 1 to 7 are required Page Number: 73-75.
to implement. 6. Barry Cipra: The Best of the 20th Century: Editors Name Top 10
Algorithms SIAM News, Volume 33, Number 4, page 1, 2000.
7. Benichou, M., J. M. Gauthier, G. Hentges, G. Ribi`ere. 1977. The
Simplex Table 3. efficient solution of large scale linear programming problems. Some
Cj (Coefficient of 3000 2000 0 0 R algorithmic techniques and computational results. Math. Programming
13 280–322.
Objective function) => k 8. Achterberg, T. 2009. SCIP: solving constraint integer programs. Math.
BV PPU SV Pj P1= P2= P3= s1 P4= s2 Programming Computation, 1 (1) 1–41.
(Cb) P0= => x1 x2 9. Bartels, R. H., G. H. Golub. 1969. The simplex method of linear
programming using LU decomposition. Communications of the
mk Association for Computing Machinery 12 266–268.
P1= 3000 m1= 1 0 1/3 -2/9 10. Harris, P. J. J. 1974. Pivot selection methods of the devex LP code. Math.
x1 30 Programming 5 1–28.
11. Hoffman, A., A. Mannos, D. Sokolowsky, D. Wiegmann. 1953.
P2= 2000 m2= 0 1 -1/3 5/9 Computational experience in solving linear prog. SIAM J.11–33.
x2 15 12. Karmarkar, N. 1984. A new polynomial-time algorithm for linear
0 0-1000 -4000/ programming, Combinatorica 4 373–395.
Cj-Zj 13. Van Roy, T. J., L. A., Wolsey. 1987. Solving mixed integer programming
/3 9 problems with automatic reformulation. Operations Research 35 (1) 45–
Conditions to reach up to the optimality: Cj-Zj≤0 for 57.
14. Mihai Banciu: Dual Simplex, School of Management Bucknell Univ.
maximization problems and Cj-Zj≥0 for minimization Lewisburg, PA 17837 [email protected] Jan 2010.
problem. 15. Dimitris Bertsimas and John Tsitsiklis. Introduction to Linear
Here all the values of Cj-Zj are either 0 or ≤0, so it’s reached Optimization. Athena Scientific, Belmont, MA, 1st edition, 1997.
16. George B. Dantzig. Maximization of a linear function of variables subject
up to the optimality. to linear inequalities. In T. C. Koopmans, editor, Activity Analysis of
x1=30, x2=15 and Max Z=1, 20,000 Production and Allocation, pages 359–373. John Wiley and Sons, Inc.,
New York, NY, 1951.
IX. CONCLUSION 17. V. Klee and G. J. Minty. How good is the simplex algorithm? In O.
Shisha, editor, Inequalities - III, pages 159–175. Academic Press, New
LPP is a method of allocating resources in an optimal way York, NY, 1972.
18. William S. Jewell: A Primal-Dual Multi-Commodity Flow Algorithm,
and it is widely used tool in operation research as a decision Operations Research Center University of California, Berkeley,
making aid in almost all industries. It refers to a planning September 1966, ORC 66-24
process that allocates resources in the best possible way so 19. Beale, E.M.L., "An Alternative Method for Linear Programming," Proo.
Cambridge Philos. Soc. 50 (l954), 513-523.
that costs are minimized or profits are maximized. The 20. Charnes, A., Cooper, W.W., and Farr, D. and Staff, "Linear Programming
simplex method is the most common way to solve large LPP. and Profit Preference Scheduling for a Manufacturing Firm," Jour. Oper.
We briefly presented seven steps involved in using the Res. Soc. Amer., _1, no. 3 (1953), 114-129.
21. Dautzig, G.B., "Maximizing of a Linear Function of Variables Subject to
simplex method and these steps will give us a general Linear Inequalities," Chapter XXI of Activity Analysis of Production and
overview of the procedure. Allocation, edited by Koopmans, T-C, Wiley, N.Y., 1951.
The simplex method is made of KE’s, Quick review can be 22. Dantzig, G.B., "Computational Algorithm of the Simplex Method", the
RAND Corporation, Paper 394, 10 April 1953. (RM-1266, Part Xll, 26
apply over seven steps algorithm to reduce the complexity October, 1953)
over computation, after 1st iteration the columns for the 23. Ford, L.R., and Fulkerson, D.R., "A Simple Algorithm for Finding
non-basic and leaving variables change. Values should be Maximal Netv.-ork Flows and an Application to the Hitchcock Problem,"
The RAND Corporation, Paper 743, 26 Sept. 1955j RIJ-1604, 29 Dec.
moved for all other RV directly into the new tableau. When 0 1956.
is found in the key column, the row always will be same in the 24. Ford, L.R., and Fulkerson, D.R., "A Primal Dual Algorithm for the
new tableau and vice-versa. In this paper simplex method Capacitated Hitchcock Problem," The RAITO Corporation, RI.i-1798
was used for solving maximization problem with constraints Part XXXIV, 25 Sept. 1956. (Naval Res. Legist. Quarto, 4, no. 1 (l957),
47-54, RAND Conference on Computational Aspects of Linear
of the form of ≤, ≥ and = constraints. We have also explored Programming, 30-31 Aug. 1956.)
the use of LPP as an economic tool for sensitivity analysis to 25. Goldman, A. J., and Tucker, A.W. , "Theory of Linear programming,"
reduce the complexity over computation using introduced Paper 4 in Linear Inequalities and Related Systems, Princeton Univ.
Press, Princeton, 1956, 53-98.
simplex algorithm. 26. S.Mittal,APande, LWang, PKumar (2011) Design Exploration and Imp
lementation ofSimplex Algorithm over Reconfigurable Computing Platf
REFERENCES orms In: IEEE International Conference on Digital
Convergence 204-209
1. T. Kitahara and S. Mizuno: A Bound for the Number of Different Basic 27. Cui, T. Ho, T. Tellambura, C. Linear Programming Detection and
Solutions Generated by the Simplex Method, Technical Report, October Decoding for MIMO Systems IEEE International Symposium on
15, 2010, to appear in Mathematical Programming (available online at Information Theory, pp. 1783-1787, July 2006.
http://www.springerlink.com/content/103081/)
2. T. Kitahara and S. Mizuno: An Upper Bound for the Number of Different
Solutions Generated by the Primal Simplex Method with Any Selection
Rule of Entering Variables.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E1948017519/19©BEIESP 42 & Sciences Publication
 Effect of Seven Steps Approach on Simplex Method to Optimize the Mathematical Manipulation

28. Majumdar, A. FPGA Implementation of Integer Linear Programming

Accelerator International Conference on Systemics, Cybernetics and
Informatics, (ICSCI), Jan 2006.
29. R.G. Bland, D. Goldfarb, and M.J. Todd, “The ellipsoid method: a
survey,” Operations Research 29, 1039-1091 (1981
30. Karmarkar, N. A new polynomial-time algorithm for linear programming
Combinatorica 4, 373−395 (1984)
31. Borgwardt, K. H. Some distribution independent results about the
asymptotic order of the average number of pivot steps in the simplex
method Mathematics of Operations Research, vol.7, no.3, pp.441-462,
1982.

AUTHORS PROFILE

Dr. MOHAMMAD RASHID

HUSSAIN, Assistant Professor,
Department of Information Systems,
College Of Computer Science, King
Khalid University, Abha, Kingdom of
Saudi Arabia, received his Master of
Technology degree from the Department of Computer Science &
Engineering, Anna University, Chennai, India. After that he
obtained his PhD degree from Bihar University, India. He was an
Associate Professor in the Department of Computer Science &
Engineering, ABESIT Ghaziabad, India. He is currently working as
an Assistant Professor in the department of Information Systems,
King Khalid University, Abha, Saudi Arabia. His research interests
include Computer Networks, Information Technology.

MOHAMMED QAYYUM, Lecturer,

Department of Computer Engineering,
College Of Computer Science, King Khalid
University, Abha, Kingdom of Saudi Arabia,
Faculty of Computer Engineering, King
Khalid University, Saudi Arabia. He worked as an Assistant
Professor under the Department of Computer Science and
Engineering at MJCET, Hyderabad, India. Prior to that, He also
worked as a Systems Engineer with designation as Member
Technical Staff for Ikanos Communications Inc. Bangalore, India.
He completed his Bachelor of Technology degree in Computer
Science & Engineering and Master of Technology degree in
Software Engineering from JNT University, Hyderabad, India. He is
keen and desperate for research in MANETS, IoT and Operation
Research.
MOHAMMAD EQUEBAL HUSSAIN,
Ph.D Scholar, Department of Computer
Science, Suresh Gyan Vihar University,
Jaipur, India, received his Master of
Technology degree from the Department of
Computer Science, Indian Statistical
Institute, Kolkata, India. He worked as an
Assistant Professor under the Department of Computer Science and
Engineering at GCET Greater Noida, India. Prior to that, He also
worked as a Software Engineer with HCL Bangalore, India. He
completed his Bachelor of Technology degree from NIT Patna,
Bihar, India. Presently he is a Ph.D Scholar in Suresh Gyan
ViharUniversity,Jaipur,India.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E1948017519/19©BEIESP 43 & Sciences Publication
View publication stats