Volume 4, Issue 3, March – 2019 International Journal of Innovativ Science and Research Technology

ISSN No:-2456-2165

Feasible Region Contraction Interior-Point Algorithm

(FERCIPA) Solver for Multi-Objective Linear
Programming Problems
Edwin F. Nsien 1, Ubon A. Abasiekwere2*, Paul J. Udoh 3
1,2
Department of Mathematics and Statistics, University of Uyo, Uyo, Nigeria
3
Department of Mathematics, University of Michigan, Ann Arbor, USA

Abstract:- This paper presents FERCIPA solver for while others are based on the interior-point algorithms, e.g.
linear programming problems. The solver which can MOSEX [2, 3]. Even though the simplex algorithm and its
handle both single objective and multi-objective linear variants have the software based on them, it has enjoyed a
programming problems of large scales generates a general acceptance and usage in solving linear
sequence of interior feasible points that converge at the programming problems. They solve linear programming
optimal solution for single objective linear problem in exponential time. An algorithm that solves
programming problems and an optimal compromise linear programming problem in polynomial time was the
solution for multi-objective linear programming interior-point algorithm developed by Karmarkar [4]. Since
problems. The solver is validated by its application to then, there has been a growing interest in the interior-point
handle single objective linear programming problems method for solving linear programming problems [5-7].
and multi-objective linear programming problems
involving up to six bounded variables and functional The method of solving large-scale linear programming
constraints. The solution obtained by FERCIPA solver problems by the interior-point method under MATLAB
is seen to compare favourably with those of other environment was presented by Zhang [8]. The existing
software like the Feasible Region Contraction interior-point algorithms have some drawbacks such as
Algorithm (FRCA) and MATLAB. extensive calculation requirements, a large number of
iterations and large computer space requirements [9]. The
Keywords:- FERCIPA Solver, Multi-objective linear methods of solving large multi-objective linear
programming, Interior feasible point, Optimal compromise programming problems by interior-point method were
solution. presented by Pandian and Jayalakshmi and Stanimirovic
etal [10, 11].
I. INTRODUCTION
In this paper, we present a software called FERCIPA
The history of linear optimization goes back to 1827 SOLVER capable of solving both single objective and
when Fourier solved the problem of finding solution of a multi-objective linear programming problems. This new
system of linear inequalities. His method was based on software provides the computerized implementation of
elimination of variables and he made an n3 algorithm to Feasible Region Contraction Interior-point Algorithm
find a feasible point or states that there is no feasible point (FERCIPA).
when there is no feasible solution to the problem. This
procedure was forgotten and rediscovered by Dines in 1918 The user interface for FERCIPA SOLVER application
and by Motzkin in 1936. This algorithm became Fourier- consists of a menu bar, an objective function data panel and
Motzkin algorithm which actually should be Fourier-Dines- a constraint equation data panel. The menu bar provides the
Motzkin and it is similar to Gaussian elimination [12]. user with access to the various application functions and
However, the algorithm is slow compared to interior point commands. The objective function data panel displays the
techniques. This method remained important even after the objective function(s) that have been specified for the
development of the simplex method since it is capable of current linear programming model. It provides command
stating the existence or nonexistence of a feasible point and buttons that allow the addition of new objective function,
also gives all the optimal solutions of a problem in integer the editing of a selected objective functions and the
linear programming [13]. removal of a specified objective function. The constraints
data panel displays the constraint equations that have been
Linear programming is, without doubt, the most specified for the current LP model. It provides command
popular tool used in an operations research study. This is buttons that allow the addition of new constraint equations,
attested to by the number of computer software that are the editing of a selected constraint equation and the
available for solving linear programming problems. Some removal of a specified constraint. At the bottom of the user
of these computer programs are based on the Simplex interface is a solve command button which the user can use
algorithm and its variants, e.g. EXCEL, MATLAB, to prompt the application to attempt the solving of the LP
LINGO, CPLEX, TORA, Optimizer in Corel Q Pro, etc. [1] problem.

IJISRT19MA523 www.ijisrt.com 738

Volume 4, Issue 3, March – 2019 International Journal of Innovativ Science and Research Technology
ISSN No:-2456-2165

The FERCIPA solver has features that include:  Step 5:

Solving of both single and multi-objective linear  Remove an objective function.
programming problems; automated generation of the Select an objective function in the objective
corresponding weights attached to multi-objective functions function panel by clicking on it, then click on the Remove
for a given problem; loading and saving of data associated button. Confirm the removal of the function by selecting
with the given linear programming problem to a storage Yes in the confirmation dialog box. Note that this operation
device; modification of data values as required; load cannot be undone.
function that allows the loading of previously saved linear
programming model data from a storage device; the  Remove a constraint.
specification of any number of decision variables for a Select a constraint in the constraints panel by clicking
given linear programming problem, saving of the solutions on it, then click on the Remove button. Confirm the
to a given linear programming problem, generating and removal of the constraint by selecting Yes in the con-
printing of linear programming problem solutions. firmation dialog. Note that this operation cannot be undone.

II. PROCEDURE FOR IMPLEMENTATION  Step 6: Solving a model

To solve the LP model, click on the LP Model menu
Here, we provide a step by step procedure for setting item and select the Solve option or click on the Solve
up and solving of single and mul-ti-objective linear button at the bottom of the screen. The application will
programming problems using the FERCIPA solver. The attempt to solve the model and a report will be generated.
solver is the computer software that provides the
computerized implementation of FERCIPA.  Step 7: Saving a model
An LP model that has been created in the application
 Step 1: Set up a linear programming problem (single can be saved for later use. To do this, click on the File
objective or multi-objective) menu item and select the Save option. In the dis-played
dialog box, specify a file name and, optionally, select a
 Step 2: Click on the LP Model Menu item and select the location for the file. Click on the save button.
New Option.
 Step 8: Loading a model
 Step 3: In the displayed dialog window, To load a previously saved model, click on the File
 input the number of decision variables in the model menu item and select the Load op-tion. In the displayed
and click on the Ok but-ton dialog box, browse to the folder containing the saved file,
 To create an objective function, click on the Add select the file, click on the Open button and the model will
button associated with the objective function panel, be loaded from the file.
input the coefficients for each decision variable in the
objective function and click on the Ok button. The III. APPLICATION OF THE FERCIPA SOLVER
entered function will then be displayed in the
objective function panel. Repeat the above steps to Consider the following single objective linear
add other objective functions. programming problem:
 To create a constraint, click on the Add button Maximize Z  25x1  7 x2  24x3
associated with the constraint panel. In the displayed s. t.
dialog, input the coefficients for the decision variable
and the right-hand value. Then select the appropriate 3x1  x2  5x3  8000000
equality sign and Click on the Ok button. The entered 5x1  x2  3x3  5000000
equation will be displayed in the constraints panel. x1 ,x2 ,x3  0
Repeat the above steps to add other constraints.
The detailed report (output) produced by the
 Step 4:
FERCIPA solver for single objective linear programming
 Edit objective function
problem is shown below and summarized in Table 1. The
Select an objective function in the objective function
CPU time in this study was gotten from a Windows PC
panel by clicking on it, then click on the Edit button. A
with 2GHz Intel processor and 2GB of RAM.
dialog box with the values of the selected objective
function will be displayed. Edit as required and click on the
 LP Problem
Ok button to update the model.
Maximize Z =25X 1  7 X 2  24X 3
 Edit constraints
Select a constraint in the constraints panel by clicking Subject to: 3X 1  X 2  5X 3  8000000
on it, then click on the Edit button. The dialog box with the 5X 1  X 2  3X 3  5000000
values of the selected constraint will be displayed. Edit as X i  0, i  1,2, 3
required and click on the Ok button to update the model.

IJISRT19MA523 www.ijisrt.com 739

 Volume 4, Issue 3, March – 2019 International Journal of Innovativ Science and Research Technology
ISSN No:-2456-2165

 Iteration 1 x*  0,500000,1500000
Z1  39500000
X1 X2 X3 Z1
Z1 :
0 500000 1500000 39500000  Time Taken
X* 0.125 seconds

 Solution

Problem Decision variable Objective function Number of iterations CPU time (seconds)
value
Linear Programming (0, 500000, 1500000) 39500000 1 0.125
Table 1:- Summary of Results Using Fercipa on Single Objective Lp Problems.

Consider the following multi-objective linear  Iteration 1

programming problems: X1 X2 Z1 Z2 Z3
*
Z1 : X 8 12 52 12 68
Molp 1 : Max z1  2x1  3x2
*
Z2 : X 15 0 30 45 60
Max z2  3x1  x2 *
Z3 : X 10 10 50 20 70
Max z3  4x1  3x2
s.t. x1  x2  20 Weights
2x1  x2  30 W1 W2 W3
x2  12 0.4333 0.2 0.3667
xi  0
x*  10.1333,8.8667 
Molp2 : Max z1  2x1  x2  3x3  x4
 Iteration 2
Max z2  x1  3x2  2x3  1.5x5
X1 X2 Z1 Z2 Z3
Max z3  2.5x1  4x2  1.5x3  3x6 Z1 :
10.3067 9.3867 48.7733 21.5333 69.3867
subject to : X*
x1  2x2  3x3  x4  30 Z2 :
15 0 30 45 60
X*
2x1  x2  x3  3x5  35
Z3 :
x1  x2  x4  x6  16 X*
10.3067 9.3867 48.7733 21.5333 69.3867
xi  0, i  1,2,...,6.
 Solution
The detailed report (output) produced by the FERCIPA x*  10.3067,9.3867 
solver for multi-objective linear programming problem 1 is
Z1  48.7733
shown below and summarized in Table 2.
Z 2  21.5333
 MOLP Problem 1 Z 3  69.3867

Maximize Z1  2X 1  3X 2  Time Taken

Maximize Z2  3X 1  X 2 0.124 seconds
Maximize Z3  4 X 1  3X 2 The detailed report (output) produced by the FERCIPA
solver for multi-objective linear programming problem 2 is
Subject to: shown below and summarized in Table 2.
X 1  X 2  20
2X 1  X 2  30  MOLP Problem 2
Maximize Z1  2X 1  X 2  3X 3  X 4
X 2  12
Maximize Z 2  X 1  3X 2  2X 3  1.5X 5
X i  0, i  1, 2
Maximize Z3  2.5X 1  4X 2  1.5X 3  3X 6

IJISRT19MA523 www.ijisrt.com 740

Volume 4, Issue 3, March – 2019 International Journal of Innovativ Science and Research Technology
ISSN No:-2456-2165

Subject to: X i  0, i  1,2,...,6

X 1  2X 2  3X 3  X 4  30
2X 1  X 2  X 3  3X 5  35
X 1  X 2  X 4  X 6  16

 Iteration 1

X1 X2 X3 X4 X5 X6 Z1 Z2 Z3
Z1 :
15.1667 0 4.6667 0.8333 0 0 45.1667 24.5 44.9167
X*
Z2 :
0 15 0 0 6.6667 1 15 55 63
X*
Z3 :
0 15 0 0 0 1 15 45 63
X*

Weights

W1 W2 W3
0.3748 0 0.6252

x*  5.6842, 9.3782, 1.749, 0.3123, 0, 0.6252

 Iteration 2

X1 X2 X3 X4 X5 X6 Z1 Z2 Z3
*
Z1 : X 13.4604 7.6981 0.3812 0 0 -5.1585 35.7623 37.3169 49.5396
*
Z2 : X 6.153 9.847 1.3843 0 3.8209 0 26.306 44.194 56.847
*
Z3 : X 0 3.694 7.5373 0 7.4402 12.306 26.306 37.3169 63

Weights

W1 W2 W3
0.429 0.2696 0.3014

 Solution
x*  7.4334, 7.0708, 2.8083, 0, 3.2725, 1.4958
Z1  30.3626
Z 2  39.1713
Z 3  55.5666

 Time Taken
0.125 seconds

IJISRT19MA523 www.ijisrt.com 741

 Volume 4, Issue 3, March – 2019 International Journal of Innovativ Science and Research Technology
ISSN No:-2456-2165

Problem Decision variable Objective function value Number of CPU time

iterations (seconds)

Z1 Z3 Z2
MOLP 1 (10.3067,9.3867) 48.7733 21.5333 69.3867 2 0.124
MOLP 2 (7.43,7.07,2.81,0, 30.3626 39.1713 55.5666 2 0.250
3.27,1.50)
Table 2:- Summary of Results Using Fercipa on Multi-Objective Lp Problems

IV. COMPARISON OF FERCIPA WITH FRCA AND comparison are the objective function value, values of
MATLAB decision variables and number of iterations. Table 3 shows
the comparison of the three methods using the single-
The performance of FERCIPA software is compared objective and multi-objective linear programming problems
with the popular interior-point algorithm operating under presented in this work.
FRCA and MATLAB environment. The modes of

Mode FERCIPA FRCA MATLAB

Single Objective Linear
Programming

Decision variable (0, 500000, 1500000) (0.0011, 499999, 1500000)

Objective function z = 39500000 z = 39499999.99
No. of iterations 1 12
Multi-objective-MOLP 1

Decision variable (10.3067, 9.3867) As known from

Objective function Z1 = 48.77, Z2=21.53, Z3=69.38 experiment
No. of iterations 2 performed
Multi-objective-MOLP 2

Decision variable (7.43,7.07,2.81,0,3.27,1.5) As known from

Objective function Z1 = 30.36, Z2=39.17, Z3=55.56 experiment
No. of iterations 2 performed
Table 3:- Comparison of the Three Software

V. CONCLUSION REFERENCES

Terlaky and Boggs [9] showed the interior point [1]. H.A. Taha, Operations Research. An
algorithm are more efficient than the simplex algorithm Introduction. New Delhi, Prentice Hall of India,
when applied to a large-scale linear programming problem, 2006.
but less efficient when applied to a small-scale linear [2]. R. Fourer, Survey of linear programming
programming problem. As it can be seen in the above table,
software. OR/MS Today, 42(3) 2001; 58-68
FERCIPA performs better than FRCA for single objective
linear programming problem and better than MATLAB for [3]. E.O. Effanga and I.O. Isaac, A Feasible Region
multi-objective linear programming problems. Hence Contraction Algorithm (FRCA) for Solving
FERCIPA by implication performs better in single, multi- Linear Programming Problems. Journal of
objective and large-scale linear programming problems. Mathematics Research, 3(3) (2011), 159-166.
[4]. N. Karmarkar, A New Polynomial-Time
The FERCIPA solver presented in this work is an Algorithm for Linear Programming,
efficient software capable of solving both single and multi- Combinatorica, 4, (1984), 373-395.
objective large-scale linear programming problems with [5]. T. Kosaki, An Interior-point method for
certain restrictions. It is programmed to solve linear minimizing the sum of piecewise-linear convex
programming problems only. The solver, in finding the
functions. Department of Industrial and
optimal solution, generates a sequence of interior-points
which converge at the optimal solution for the single- Management, Tokyo Institute of Technology,
objective linear programming problem and generates a (2010), 1-9.
sequence of weights attached to multi-objective linear [6]. M. Ehrgott, L. Shao and A. Schobel, An
programming problem. Approximation algorithm for convex

IJISRT19MA523 www.ijisrt.com 742

Volume 4, Issue 3, March – 2019 International Journal of Innovativ Science and Research Technology
ISSN No:-2456-2165

multiobjective programming problems. Journal of

Global Optimization, 50 (2011), 397-416.
[7]. M.L. Bougnol, J.H. Dula and P. Rouse, Interior-
point methods in DEA to determine non-zero
multiplier weight. Computer and Operations
Research, 39 (2012), 698-708.
[8]. Y. Zhang, Solving large scale linear programming
by interior point methods under MATLAB
environment, Technical Report, Mathematics.
Department of Maryland Baltimore Country,
1996.
[9]. T. Terlaky and P.T. Boggs, Interior-point
methods. In: Encyclopedia of Operations
Research and Management Science. Springer,
Boston, MA, 2001.
[10]. P. Pandian and M. Jayalakshmi, Determining
Efficient Solution to Multiple Objective Linear
Programming Problems, Applied Mathematical
Sciences, 7(26) (2013), 1275-1282.
[11]. I.P. Stanimirovic, M.L. Zlatanovic and M.D.
Petkovic, On the linear weighted sum method for
multi-objective optimization, Facta Acta
University, 26 (2011), 49-63.
[12]. P. Parrilo and S. Lall, Fourier-motzkin
elimination, Stanford University, (2003), 1–29.
[13]. G. B. Dantzig and B. C. Eaves, Fourier–motzkin
elimination and its dual, Journal of Combinatorial
Theory, (A) (1973), 288–297.

IJISRT19MA523 www.ijisrt.com 743