Mathematical Statistician and Engineering Applications

ISSN: 2094-0343
2326-9865

Solution of Unconstrained Non-Linear Programming Problems using

Differential evolution, Genetic algorithm and Artificial bee colony
evolutionary algorithms: A comparative study

Priyavada#1, Binay Kumar *2

1, 2
Department of Mathematics, Lingaya’s Vidyapeeth, Faridabad (Delhi NCR)
1
[email protected]
2
[email protected]
*Corresponding author: Dr. Binay Kumar

Article Info Abstract

Page Number: 1003 - 1019 Evolutionary computation (EC) is a set of global optimization algorithms
Publication Issue: inspired by biological evolution. EC techniques, which are based on
Vol 72 No. 1 (2023) principle of evolution and survival of the fittest Darwinian Theory. It is a
sub-field of Artificial Intelligent techniques and soft computing. In
Evolutionary computation, we use non classical methods based on natural
evolution instead of classical methods. When fitness function has several
local extreme or not known derivative then it is very difficult to use
classical methods for solution of optimization problem in such situation
non classical methods gives very convenient and faster solution. In this
paper we present Genetic algorithms, Artificial bee colony algorithm and
Differential evolution algorithm of Evolutionary computation technique
for solution of unconstrained non-linear programming problem. Numerical
examples have solved using software MATLAB and the result has
compared among evolutionary algorithms as Genetic algorithm,
Article History differential algorithm and artificial bee colony algorithm.
Article Received: 22 October 2022
Keywords: Evolutionary computation, Genetic algorithm, Differential
Revised: 20 November 2022
evolution algorithm and Artificial bee colony algorithm.
Accepted: 24 December 2022

1. Introduction
Last few decades there are a lot of evolutionary algorithms subset of evolutionary computational
approach have been proposed and used to many real world complex optimization problems and got
the success in finding the near best optimal solution. Solution of non-linear optimization problems
is one of the most important issues in many field like complex research problem and real world
problems. Evolutionary algorithm uses the computational model of evolutionary processes as key
element and share a usual concept based on the processes of selection and reproduction. Mostly all
evolutionary algorithms are inspired by the natural concept of Darwinian Theory and this theory
tells that EAs are not static but dynamic as they can evolve over time. Evolutionary algorithms are
population based local search algorithm. During the last twenty years most of the research has done
in the complex objective function without using constraints. Genetic algorithm, differential
evolution, Particle swarm optimization, Artificial Bee Colony algorithm, Ant colony optimization

1003
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

and Spider monkey optimization etc. are the evolutionary algorithms and all type of non-linear
complex optimization problem can be easily solved by these algorithms. Among all Differential
evolution is very easy and effective to apply. GA is evolutionary algorithm based on the theory of
natural selection of Charles Darwin. GA and DE both uses same operator’s mutation, crossover and
selection but in GA crossover is used first then mutation while in DE mutation is used first then
crossover. DE algorithm solved continuous optimization problems first and depends on the
differentials of two vectors. ABC algorithm introduced by Dervis Karaboga in 2005. It is inspired
by the behaviour of natural honey bees and uses three main phases employed bee phase, onlooker
bee phase and scout bee phase. In ABC algorithm the aim of the bees is to find the best place of
food sources with high nectar amount.
2. Literature Review
2.1 Genetic algorithms
Nuno Neves et al. [1] discussed the study of NON-LINEAR Optimization problem using a distributed GA.
GA is used in large space for the optimization problem. Dr. Yuyi Lin et al. [2] proposed solution of non-
linear optimization problems using GA with penalty method. This paper discussed the techniques of applying
GAs with penalty function to solve highly constrained nonlinear engineering design problems. Subrata Datta
et al. [3] proposed the efficient genetic algorithm on programming problems for fittest chromosomes.
Chhavi Mangla et al. [4] proposed the genetic algorithm based optimization problem for solution of
system of non-linear equations. D.E technique is applied for the solution of non-linear optimization
problem. This type of problems is changed first in to multi objective problem then a new fitness
function is used further process and both result is compared. Dr. Amanpreet Singh et al. [5]
discussed some special type of Unconstrained Optimization Problem with using G.A and main
focus of G.A is on the simulation of natural evolutionary process which can use for certain standard
optimization problems. It is evolutionary algorithm which works according to the basis of natural
genetics on a population string structures. The global and population based optimization technique
is used to avoid the problem like stagnation of the solution to a local minimum. Punam S Mhetre et
al. [6] introduced G.A for the solution of both linear and Non-linear optimization problems and
check the major benefits acquired as a result of using G.A. Gauss- Legendre integration method as a
technique is used for the solution of NLPP and to get the result without changing the non -linear
equation to linear. G.A result are very good for all tasks requiring optimization and highly
successful in any situation where many variables interact to produce a large number of possible
solutions. T Yokota et al. [7] discussed non- linear optimization problems of integer programming
and also its applications. T. Yokota also introduced the new method for the solution of NLPP for
better comparison. Some methods and holding property are discussed with the help of G.A. From
the result it is observed that proposed method is best then the property. Chhavi Mangla et al. [8]
discussed solution of non- linear optimization problem by using Genetics algorithm. Both type of
problems singles and multi-objective optimization problem based on engineering, social sciences
and medical sciences are solved by using standard benchmark problems. This paper carries out a
comprehensive review to analysis the work of GA.

1004
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

2.2 Artificial Bee Colony Algorithm

Lalit Kumar et al. [9] proposed solution of NP-hard Problem using ABC algorithm. ABC is considered as a
newest nature-inspired swarm intelligence-based optimization algorithms. S. T S. Talatahari et al. [10]
discussed solution of Parameter Identification of Nonlinear Problems by Artificial Bee Colony Algorithm.
Dervis Karaboga et al. [11] discussed the ABC algorithm for solution of non -linear unconstraint
optimization problem. Modified result of ABC algorithm has been compared to state of art
algorithm. Result of modified ABC algorithm are tested for real engineering problems and found
that it can efficiently applied for solving constrained optimization problem. Weifang Gao et al. [12]
introduced a framework of constraint optimization problem by novel ABC algorithm. Bi -objective
optimization problem is taken in which first part remains same (objective function) and the second
part is the degree of constraint. Multi-strategy technique is used which consists three types of
strategies, and play the main role between diversity and convergence. Soudeh Babaeizadeh et al.
[13] discussed Constrained based ABC Algorithm for Optimization Problems. It is an optimization
algorithm and compared to other optimization based algorithms. ABC has one poor exploitation
ability that can be removed by using three types of phases employee bees, onlooker bees and scout
bees phase by a new search technique. Soudeh Babaeizadeh et al. [14] discussed non-linear
optimization problem related to enhanced Constrained ABC Algorithm. From last few years ABC
with less parameters and strong global optimization ability researcher paid their attention on these
two problems. But still there is one drawback in ABC algorithm that is poor in exploitation.
Numerical examples results are shown with several benchmark functions and result are compared.
Nadezda Stanarevic et al. [15] proposed improved ABC algorithm for constrained optimization
problem. Artificial bee colony algorithm is applicable only on unconstrained optimization problems
but later a lot of modification in ABC was introduced of constraints problems. ABC modified
version introduced for constraint optimization problem.
2.3 Differential Evolution Algorithm
Jouni Lampinen et al. [16] discussed the mixed variable non- linear optimization by differential
evolution. This paper solving non-linear programming problems containing integer, discrete and
continuous variables. Md. Abul Kalam Azad et al. [17] proposed the solution of non -linear
continuous global optimization problems by modified differential evolution. This paper discussed
about modified differential evolution in which some self -adaptive parameter is introduced.
Mutation is modified and then use its inversion operator. Some non-linear continuous optimization
problems are tested. Inversion operator has used in this proposed differential evolution algorithm.
Musrrat Ali et al. [18] discussed simpler differential evolution algorithm. Simple and modified
version of differential evolution algorithm (NSDE) is presented. In simple DE & NSDE there is
only one difference of initial population between them. NSDE uses non -linear simplex method to
create the initial population. Twenty benchmark problem are taken with box constraint and
numerical result are compared with traditional DE and other DE. Zhang Xiao Fei et al. [19]
proposed different concept related to reactive power optimization of power system based on niching
differential evolution algorithm, As the problem is non -linear integer type. In this Zhang introduced
the niche theory and some improvement in the differential evolution which aim of the modified
RPO. Local search ability and broaden search range also increased. Different degrees of
improvement are also used for speed and precision of convergence. Chiha Ibtissem et al. [20]

1005
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

discussed differential evolution algorithm for special multi -objective optimization problems. A new
variant is discussed for DE. Proposed algorithm named based cost differential evolution generate
the mutant vector. This vector is created by adding a weighted difference of two cost functions.
Then performance of Bc.DE is calculated on several bound constrained non -linear and non -
differentiable numerical problems and compare with the DE and other DE variants. Jeerayut
Wetweerapong et al. [21] Proposed an improved differential evolution algorithm with a restart
technique to solve system of non -linear equations, which often used for the solution of complex
computational problems with variables. In DE –R a new approach is used for mutation and restart
technique to stop premature convergence. Proposed technique is applied on various world and
synthetic problems and compare with other methods, it found that DER gives fast convergence and
high quality solutions. Saeed Nezhadhosein et al. [22] Introduced integrating differential evolution
algorithm with modified hybrid GA for solving non -linear optimal control problems. Two phase
algorithm is used based on IDE. Associate non- linear problems with hybrid GA of optimal control
problem is discussed. In First phase DE is completely depends on initial population. In 2nd phase
MHGA starts with new population.
3. Evolutionary Computational Approach
3.1 Genetic Algorithm
Genetic Algorithms are metaheuristic optimization method and based on the biological theory
introduced by Charles Darwin and provides an extended concept survival of the fittest [23]. GA is
developed by Professor Johan Holland along with his students and colleagues at the University of
Michigan during the 1960s and 1970s. [24] Dr Johan Holland then developed the theoretical
framework for applying genetic algorithm many complex computational problems needed searching
through huge search domain. In that cases G.A gives best and efficient tool to reach at an optimal
solution and gives an efficient way for solving computationally hard non-linear optimization
problems. GA starts with a set of possible solutions where each solution in GA, is represented as a
chromosome. Set of operators are applied to the initial set of solution space. GA does not work on a
single trial solution at a time but work on entire population. In special generation, the whole
population comprises to the set of solution then select the fittest member and share their qualities.
GA is based on the principal of Natural selection and applied to discrete optimisation problem. It is
frequently used to find optimal or near optimal solution to difficult problem. GA can solve both
types of problems unconstrained and constrained optimization problems [25].
GA has three main operators:
3.1.1 Selection Operator
This operator gives preference to better solution allowing them to pass on their genes to the
next generation of the algorithm. The main objective is to select the good solution and discard the
bad solution. The individual with a higher fitness must have a higher probability of having
offspring. There are different selection techniques in GA.
(a) Roulette Wheel Selection. Here selection is done by Roulette wheel.
(b) Rank Based Selection: Rank the all individual firstly and then give the fitness ranking to each.

1006
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

(c) Tournament Selection: A group of n individuals is randomly selected from the current population
and from them best is selected. Here n is the tournament size the higher n the higher the pressure to
select above average quality individual. A lot of tournaments are held between the individual and this
process is repeated as often as desired.

3.1.2 Crossover Operator

The crossover between two parents string produce offspring by swapping genes of the
chromosomes. With the help of crossover, we can combine partial solution from different
candidate.
In GA three types of crossover are used
1-point Crossover,
2-point crossover and
3-Uniform- Crossover.

3.1.3 Mutation Operator

GA get better solution by using this operator. This operator is used to maintain genetic diversity
from one generation of population to the coming generation. After mutation the solution may
change entirely from the last solution. Mutation operator changes 1 to 0 or 0 to 1.
Bit Flip Mutation: We can select one or more random bits and flip them to each other
Swap Mutation: In swap mutation we select two position on the chromosome at random and
interchange the values
Scramble Mutation: This is also famous with permutation representation choose, select a subset of
genes and shuffled their values.
Inverse Mutation: In Inverse mutation, we select a subset of genes and inverse that selected
subset.

3.2 Artificial Bee Colony Algorithm

ABC is same as swarm based intelligence, metaheuristic and global optimization algorithm and
advanced proposed algorithm by Dervis Karaboga in 2005. Dervis Karaboga with his research
members studied on ABC algorithm and some of its applications. The first journal article of ABC
was presented by Karaboga and Basturk in 2007 in which the performance of ABC was compared
to GA and PSO. First conference paper is published in 2006. In ABC algorithm [26, 27], the swarm
of artificial bees are categorised in three groups of bees: employed bees, onlookers and scout bee.
50% are the employed bees and 50% are onlooker’s bees to the total bees and only 1 and 2 are scout
bees. The total numbers of employed bees are equal to the number of onlooker’s bees. Every food
source is a possible solution to the optimization problems. Here the main aim of the bees is to find

1007
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

the best place of food source with high amount of nectar. Fitness of bees is determined in terms of
best quality of the food source. Initially first group of employed bees goes to random places for
searching the food. At that time onlooker bees are at bee hive. They are waiting for the employed at
the hive. When the employed returns to beehive then they share all the information like quality,
quantity and distance with onlooker bees regarding the food source. After receiving all information
from the employed bees the onlooker bees also start searching in neighbourhood of same food
source. Food sources which are abandoned changed by scout bees in to new randomly generated
food source.
Behaviour of Honey Bee in Nature
ABC algorithm is inspired by honey bees, where honey bee’s swarm works in collective
intelligence way, while searching the food. Bees have many qualities, like bees can communicate in
their own language, can memorize all the information and take best own decisions based on that
gathered information. By the intelligence behaviour of bees, many researchers motivate to simulate
above foraging behaviour of the bees. The nature of honey bees is listed as
Employed Bees: Every employed Bee exploit only single food source and store every information
regarding that food source like richness, direction and distance.
Unemployed bees: These bees divided in to two groups of bees, onlooker and scout bees. Only
these bees found and shape out all information from 1st group of bees based on certain probability.
Foraging behaviour: Foraging behaviour is the important characteristic of the bees. In this process,
bees go away from their hive and starts searching best food with high nectar amount, extracts all
nectar and stores in her abdomen. She can store nectar till 30-120 minutes, on the basis of nectar
quantity and distance. Finally, they share all information.
Four Phases of ABC Algorithm
1. Initialization Phase: In this phase we select the food sources randomly with the
expression.
𝑗 𝑗 𝑗 𝑗
𝑥𝑖 =𝑥𝑚𝑖𝑛 +rand(0,1)×(𝑥𝑚𝑎𝑥 _𝑥𝑚𝑖𝑛 )
𝑗 𝑗
Here 𝑥𝑚𝑖𝑛 & 𝑥𝑚𝑎𝑥 are the upper and lower bound of the solution space (Domain) of objective
function, rand (0 1) is a random number lies between [0 1]. And fitness function is determined by
this formula
Evaluate the fitness function of all new solution with the help of this formula.
1
𝑓𝑖𝑡 = {1+𝑓 , 𝑖𝑓 𝑓 ≥ 0 1 + 𝑓𝑎𝑏𝑠(𝑓), 𝑖𝑓 𝑓 < 0

2. Employed Bee Phase: Three steps are used mainly in this phase a) Generate a new solution
b) Calculate new fitness, c) Apply greedy selection
The neighbourhood food source is determined and calculated by the following equation.

1008
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

𝑗 𝑗
𝑥𝑛𝑒𝑤 = 𝑥 𝑗 +ɸ (𝑥 𝑗 −𝑥𝑝 ), ɸ ∈ [-1,1]

Where x = Current solution

𝑗 𝑗
𝑥𝑝 =Random partner, 𝑥𝑛𝑒𝑤 =New solution

f=Objective function value of a solution.

3. Onlooker Bee phase

Firstly, calculate the probabilities of every food source. Then generate a new solution that depends
on probability value after that Calculate new fitness by applying greedy selection
𝑓𝑖𝑡
Probability = 0.9× 𝑓𝑖𝑡𝑖 + 0.1

Here 𝑓𝑖𝑡𝑖 is the fitness value of the solution i. Onlooker bees search the neighborhood food sources
according to this expression.
𝑗 𝑗
𝑥𝑛𝑒𝑤 = 𝑥 𝑗 +ɸ (𝑥 𝑗 −𝑥𝑝 )

4. Scout Bee Phase

First of all, found the abandoned solution which based on limit value. Then discard that abandoned
solution and generate a new solution randomly to replace that. The new food sources are randomly
generated by the scout bees. The new solution will be discovered by the scout bees by using the
expression.
𝑗 𝑗 𝑗 𝑗
𝑥𝑖 =𝑥𝑚𝑖𝑛 +rand (0, 1) × (𝑥𝑚𝑎𝑥 _𝑥𝑚𝑖𝑛 )

3.3 Differential Evolution Algorithm

Differential Evolution Algorithm is very simple and easy stochastic optimization technique for the
solution of non- linear optimization problems which evolves a population of potential solutions
(individuals) in order to increase the convergence to optimal solutions. DE is population based
stochastic search algorithm developed by Storn and Price in 1995 for the solution of Chebyshev
fitting problems [28]. DE is different from other algorithm in the sense that distance and direction
from current population is used to guide the search process. DE has evolutionary operators like
mutation, crossover and selection and totally depends on difference of vectors.DE is not nature
inspired algorithm like PSO, ABC etc. But in DE global search algorithm employ investigate to
allow escaping from local minima. Here three main operators are used mutation, crossover and last
selection. The special work of DE algorithm is generating a trial vector, for this three different
vectors are selected from the initial population and got the trial vector by adding a weighted
difference of two randomly selected solutions to a third one which is target vector. This process is
known as mutation operator of evolutionary algorithm. After mutation operator vectors are merged

1009
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

with another vector and this operation is called crossover operators. After this process we get a new
trial vector and this trial is accepted after applying greedy selection in new population if its fitness
function value is better than the target vector fitness value. It evolves generation to generation until
the stopping criteria would be met. There is one difference between GA and DE algorithm is the
mutation and recombination phase. In DE position of the vectors give important information about
the best vectors. Diversity of the present population is determined by the distance between the
individuals. If distance is large then steps size are also large for the individuals and they can explore
much search space in another side if distance is small, then step size should be small to exploit local
areas. Minimum population size in differential evolution should be 4.
Three Main Operators of Differential Evolution
3.3.1 Mutation Operator
In DE trial vector plays an important role and it is created by mutation operator.
Trial vector is created for each solution as follow
Trial Vector=Target vector + Scale factor× (Randomly selected solution1- Randomly
selected solution12)
Five mutation strategies are proposed by Storn and Price initially. Later Storn and Price
discussed ten different working strategies of DE and these are derived from mutation
schemes. Every mutation strategy either used exponential type crossover or the binomial
type crossover.
DE/rand/1:Uij= 𝑥𝑟1 j+F(𝑥𝑟2𝑗 − 𝑥𝑟3 j)
DE/best/1:Uij= 𝑥𝑏𝑒𝑠𝑡 j+F(𝑥𝑟2 𝑗 − 𝑥𝑟3 )
DE/target to best/2:Uij= 𝑥𝑖 j+F(𝑥𝑏𝑒𝑠𝑡 j -𝑥𝑖 j )+F(𝑥𝑟1𝑗 − 𝑥𝑟2 j)
DE/best/2:Uij= 𝑥𝑏𝑒𝑠𝑡 j+F(𝑥𝑟1 𝑗 − 𝑥𝑟2 )+ F(𝑥𝑟3𝑗 − 𝑥𝑟4 )
DE/rand/2:Uij= 𝑥𝑟1 j+F(𝑥𝑟2𝑗 − 𝑥𝑟3 j)+ F(𝑥𝑟4𝑗 − 𝑥𝑟5 )

3.3.2 Crossover Operator

Crossover is the discrete recombination of parent and trial vector and increase the diversity.
Crossover is of two type Binomial and Exponential crossover. Binomial crossover is
commonly used by the researchers as it gives best result. Any calculation work is not used in
crossover operator only some conditions which are user defined gives the trial vector. Trial
vector is generated as
Uj = {𝑣𝑗 , 𝑖𝑓 𝑟 ≤ 𝑝𝑐 𝑂𝑅 𝑗 = 𝛿 𝑥𝑗 𝑖𝑓 𝑟 > 𝑝𝑐 𝐴𝑁𝐷 𝑗 ≠ 𝛿
Here pc is the Crossover probability
𝛿 =Randomly selected variable 𝛿 ∈ {1,2,3,4, … … , 𝐷}

𝑟 = 𝑟𝑎𝑛𝑑𝑜𝑚 𝑛𝑢𝑚𝑏𝑒𝑟 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 0 𝑎𝑛𝑑 1

Uj jth variable of trial vector
𝑣𝑗 jth variable of donor vector

𝑥𝑗 jth variable of target vector

1010
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

3.3.3 Selection operator

In this operator apply greedy selection and find the best fitness vector which can take part in
mutation operator to produce best trial vector which will continue in next iteration. It constructs the
new generation for the next iteration and apply deterministic selection. The offspring replaces the
parent if the fitness of the offspring is better than its parent otherwise the parent survive to the next
generation. This ensure that the average fitness of the population does not deteriorate.
4. Non Linear Programming Problems

There are different types of non –linear optimization problems like continuous optimization, Bound
constraint, discrete, global derivative free and non -differentiable optimization problems. Among
these problems some are solved by traditional method which give the approximate solution result
rather than accurate solution and some are solved by evolutionary algorithm which are derivative
free algorithm. As everyone know that solution of NLPP are difficult in comparison to LPP. Many
traditional methods are available for the solution of LPP and these are easy to apply. There are two
types of NLPP constraints and without constraints. NLPP without constraints could be designed
with several methods like steepest descent, Newton’s method. They are taking less time in
comparison to the NLPP with constrains. On the other hand, various methods are available like
KKT conditions, Lagrangian multiplier, penalty method, barrier method for the solution of NLPP
with constraints and there are special problems of quadratic NLPP that cab solved by Beal’s
method, Wolfe’s method etc. Special method is not available for the solution of both types of
problems LPP and NLPP. Many non-traditional methods have been introduced for the solution of
different optimization problems. These methods are very powerful, easy, populations based and
metaheuristic algorithm, by which we can solve NLPP without constrains and with constraints like
PSO, DE, ABC and GA to implement for the solution of complex research problems and others
problems. DE is very easy to apply because this algorithm used less parameters to others algorithm.
In NLPP the function may be convex or non-convex. A convex optimization problem maintains the
properties of a linear programming problem and a non-convex problem the properties of a nonlinear
programming problem. The basic difference between the two categories is that in a) convex
optimization there can be only one optimal solution, which is globally optimal or you might prove
that there is no feasible solution to the problem, while in b) non convex optimization may have
multiple locally optimal points and it can take a lot of time to identify whether the problem has no
solution or if the solution is global.
Suppose we are given an NLP with constraint
Min. f (x)
Subject to linear or non-linear constraints
g(x) ≤ 0; and h(x) = 0;
5. Numerical Examples
1.objective function (min) = x1-x2+ 2x12+2x1 x2+ x22

1011
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

0≤x1, x2≤10
2.objective function(min) = 20x1+26x2+4x1x2-4x12-3x22
-10≤x1, x2≤10
Genetic Algorithm problems
1.Objective function (min) = x1-x2+ 2x12+2x1 x2+ x22
0≤x1, x2≤10
1st Table of G.A

S.No Population No. of Variable Generation Function Value X1 X2

1 10 2 50 -0.22185 0 0.667
2 20 2 10 -0.25 0 0.5
3 50 2 50 -0.2599 0 0.5
4 100 2 50 -0.25 0 0.5
5 100 2 100 -0.25 0 0.5
6 500 2 100 -0.25 0 0.5

7 500 2 500 -0.25 0 0.5

Figure 1 (GA)

2. Objective function(min) = 20x1+26x2+4x1x2-4x12-3x22

-10≤x1, x2≤10

1012
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

2nd Table of G.A

S.No Population No. of Variable Generation Function Value X1 X2

1 10 2 50 -1072.0593 9.998 -9.292
2 20 2 10 -1144.975 9.971 -9.903
3 50 2 50 -1159.695 9.998 -9.999
4 100 2 50 -1159.999 10 -10
5 100 2 100 -1159.999 10 -10
6 500 2 100 -1159.999 10 -10

7 500 2 500 -1159.999 10 -10

Figure 2 (GA)

Artificial Bee Colony Algorithm problems

1.Objective function (min) = x1-x2+ 2x12+2x1 x2+ x22
0≤x1, x2≤10
1st Table of ABC

Function
Pop Size Variable Iteration value X1 X2 Fnew
10 2 50 -0.25(38ite) 0 0.5006 -0.2500
20 2 10 -0.2103(1it) 0 5.0616 20.5585
50 2 50 -0.25(40ite) 0.0011 0.4290 -0.2429
100 2 50 -0.25(10ite) 0 0.3876 -0.25
100 2 100 -0.25(39ite) 0 0.5000 -0.25
500 2 100 -0.25(4ite) 0 0.5000 -0.25
500 2 500 0(2iter) 0 0.5000 -0.25

1013
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

For ABC solution is -0.2103 in 10 iterations with 20 populations but when iteration and population
are increasing ABC gives the best optimal solution is -0.25 near to 0 in 50 iterations and all other
iteration with different population. Values of x1 =0 and x2=0.5.
Figure 3(ABC)

2.Artificial Bee Colony Algorithm problems

2.objective function(min) = 20x1+26x2 +4x1x2-4x12-3x22
-10≤x1 , x2≤10
2nd Table of ABC

Function
Pop size Variable Iteration value X1 X2 fnew
10 2 50 -865(50iter) -9.9823 -4.5912 -597.5201
20 2 10 -1080.337(5it) -9.9987 6.4894 -817.0251
50 2 50 -1160(20ite) 10 -8.6092 -990.5670
100 2 50 -1160(8ite) 10 -10 -1160
100 2 100 -1160(6ite) -10 -2.5203 -583.7715
500 2 100 -1160(2ite) -9.7807 10 -1.0095e+03
500 2 500 -1160(3iter) 10 -10 -1160

For ABC solution is -0.2103 in 10 iterations with 20 populations but when iteration and population
are increasing ABC gives the best optimal solution is -0.25 near to 0 in 50 iterations and all other
iteration with different population. Values of x1 =0 and x2=0.5. In the same way for 2nd problem we
got the optimal solution -1160 when we take 50 iterations with 50 populations and gives the best
solution in 20 iterations. When iterations are increasing solution obtained in less number of
iterations.

1014
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

Figure 4

3.Differential Evolution Algorithm

1.Objective function (min) = x1-x2+ 2x12+2x1 x2+ x22
0≤x1 , x2≤10
1st Table of DE

Pop Size Variable Iteration Best Cost X1 X2 New Sol

5 2 50 29.2154(54) 2.6437 2.0023 2.6437, 2.0023
20 2 10 -0.2498(3) 0.0104 0.5143 0,0.5221
50 2 50 0(1it) 0 0.5000 0,0.5000
100 2 50 -0.25(20ite) 0 0.5000 0,0.5000

In DE result is faster than ABC and GA here optimal solution is near to 0 in 3rd iteration with 20
populations and x1=0.0104, x2= 0.5143. when iterations and population are increasing than we got
optimal point in minimum no of iterations.
2.objective function(min) = 20x1+26x2 +4x1x2-4x12-3x22
-10≤x1 ,x2≤10
2nd Table of DE

Pop Size Variable Iteration Best Cost X1 X2 New Sol

5 2 50 -1.160(3ite) 10 -10 10,-10
20 2 10 -1160 10 -10 10,-10
50 2 50 -1.160 10 -10 10,-10
100 2 50 -1160(2ite) 10 -10 10,-10

1015
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

Result of the problems

In GA 1st problem we got the optimal solution -0.25 near to 0 because problem is of minimization
in 10 number of iterations with 20 population and values of x1=0, x2=0.5. These values are same
for all other iterations in the table. so we get our optimal solution in 10 number of iterations. 2nd
problem gives the optimal solution -1154.695 in 10 number of iteration with 20 populations, value
of x1 and x2 are near to 10 and -10. These value remain same for all other iteration also in the table.
For ABC solution is -0.2103 in 10 iterations with 20 populations but when iteration and population
are increasing ABC gives the best optimal solution is -0.25 near to 0 in 50 iterations and all other
iteration with different population. Values of x1 =0 and x2=0.5. In the same way for 2nd problem we
got the optimal solution -1160 when we take 50 iterations with 50 populations and gives the best
solution in 20 iterations. When iterations are increasing solution obtained in less number of
iterations. In DE result is faster than ABC and GA here optimal solution is near to 0 in 3rd iteration
with 20 populations and x1=0.0104, x2= 0.5143. when iterations and population are increasing than
we got optimal point in minimum no of iterations. In 2nd problem optimal point -1160 is obtained in
second iteration when population is 100 and iteration are 50. Value of x1, x2 are 10 and -10 So we
can conclude that DE gives faster convergence point in less iterations. Result of DE is better in
comparison of ABC and GA.
COMPARISON TABLE OF 1st problem
Pop Variable Gen GA ABC DE

X1 X2 F.V
X1 X2 F.V X1 X2 F.V
-
10 2 50 0 0.500 0.25(38ite 2.64 29.2154(
0.667 0.22185 0 6 r) 37 2 54ite)
-
20 2 10 0 5.061 0.2103(1i 0.01 0.51 -
0.5 -0.25 0 6 t) 04 43 0.2498(3)
-
50 2 50 0 0.001 0.429 0.25(40ite
0.5 -0.2599 1 0 ) 0 0.5 0 (1st)
-
100 2 50 0 0.387 0.25(10ite
0.5 -0.25 0 6 ) 0 0.5 0
-
100 2 100 0 0.500 0.25(39ite
0.5 -0.25 0 0 ) 0 0 0
0.500 -
500 2 100 0
0.5 -0.25 0 0 0.25(4ite) 0 0 0
0.500
500 2 500 0
0.5 -0.25 0 0 0(2iter) 0 0 0

1016
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

COMPARISON TABLE OF 2nd problem

No. of
Population Variabl Gen
e GA ABC DE

X1 X2 F.V
X1 X2 F.V X1 X2 F.V
- -
10 2 50 4.591 1.160(3i
9.998 -9.292 -1072.059 -9.9823 2 -865(50iter) 10 -10 te)
-
20 2 10 6.489 1080.3371(5i
9.971 -9.903 -1144.975 -9.9987 4 t) 10 -10 -1160
-
50 2 50 8.609
9.998 -9.999 -1159.695 10 2 -1160(20ite) 10 -10 -1.160
-
100 2 50 1160(2it
10 -10 -1159.999 10 -10 -1160(8ite) 10 -10 er)
-
100 2 100 2.520
10 -10 -1159.999 -10 3 -1160(6ite) 10 -10 -1160
500 2 100 10 -10 -1159.999 -9.7807 10 -1160(2ite) 10 -10 -1160
500 2 500 10 -10 -1159.999 10 -10 -1160(3iter) 10 -10 -1160

7.Conclusion
This paper discussed the solution of unconstrained non-linear optimization problems by using
evolutionary computational approaches Genetic algorithm, Artificial bee colony algorithm, and
Differential evolution by using MATLAB. For this comparison unconstrained non-linear
optimization problems are solved and result are compared. According to 1 st comparison table result
of DE at 50 number of iteration and 50 populations is 0 in 1st iteration but GA and ABC result are –
0.2599 and -0.25 in 40 number of iteration, so result is faster than both algorithms. In 2nd
comparison table we can see result is better in with 20 particles and 10 number of iterations DE
result is -1160 which is minimum but GA and ABC result are -1144.975 and -1080. So DE result is
better in comparison GA and ABC algorithm in less number of iterations. These algorithms are
computational algorithms and efficient for the solution of non-linear optimization problems. From
tables and figures it is concluded that Differential evolution gives better result in less number of
iterations. Optimal solution obtained from the DE algorithm is much better than Genetic algorithm
and ABC algorithm.

References:
[1] Nuno Neves, Anthony-Trung Nguyen, Edgar L. Torres. A study of non-linear optimization problem using a
distributed Genetic algorithm, in proceedings of the 25th International Conference on Parallel Processing,
August 1996.
[2] Dr. Yuyi Lin Yang Shen. Dealing with non-linear optimization problems using algorithm with penalty
function.

1017
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

[3] Subrata Datta, Chayanika Garai and Chandrani Das. The efficient genetic algorithm on programming
problems for fittest chromosomes, Volume 3, No. 6, June 2012 Journal of Global Research in Computer
Science, ISSN-2229-371X.
[4] Chhavi Mangla, Moin Uddin and Musheer Ahmad. G.A based optimization for system of Non –Linear
equation. International Journal of advanced Technology and
engineeringexploration,vol(5144),ISSN:2394,5443.http:dx.doi.org/10.191901/, IJATEE,2018.543009.
[5] Dr. Amanpreet Singh, Shivani Sanan , Amit Kumar. Solving Unconstrained Rosenbrock Optimization
Problem Using Genetic Algorithms at High Technology Letters. ISSN NO: 1006-6748http://www.gjstx-e.cn/,
Volume 27, Issue 3, 2021
[6] Punam S Mhetre. Genetic algorithm for the solution of both linear and Non-linear optimization problems .
International Journal of Advanced Engineering Technology, E-ISSN 0976-3945, IJAET/Vol.III/ Issue II/April-
June, 2012/114-118 (2012).
[7] T. Yokota, M. Gen. Solving non-linear Integer Programming problem using. Proceeding of IEEE International
conference on system, Man and cybernetics 5oct 1994, DOI:10.1109/ICSMC.1994.400076.
[8] Chhavi Mangla, Moin Uddin and Musheer Ahmad. Optimization of complex Non –Linear system using
Genetic algorithm. International Journal of information technology, ISSN 2511-2104, DOI 10,1007/541870-
020-00421-z ,2020.
[9] Lalit Kumar, Dr. Dheerendra Singh. Solution of NP-hard Problem using ABC algorithm. International journal
of engineering and technology. ISSN 0976 – 6367(Print) ISSN 0976 – 6375(Online) Volume 4, Issue 1,
January- February (2013), pp. 171-177.
[10] S. Talatahari, H. Mohaggeg, Kh. Najafi and A. Manafzadeh. Solving Parameter Identification of Nonlinear
Problems by Artificial Bee Colony Algorithm, Hindawi Publishing Corporation Mathematical Problems in
Engineering Volume 2014, Article ID 479197, 6 pages http://dx.doi.org/10.1155/2014/479197.
[11] Dervis Karaboga and Bahriye Basturk. Artificial Bee Colony (ABC) Optimization Algorithm for Solving
Constrained Optimization Problems. P. Melin et al. (Eds.): IFSA 2007, LNAI 4529, pp. 789–798, 2007. c
Springer-Verlag Berlin Heidelberg 2007.
[12] Weifeing Gao, Lingling Huang, Yuting Luo, Zhifang Wei, and Sanyang Liu. Constrained Optimization by
Artificial Bee Colony Framework. Digital Object Identifier 10.1109/ACCESS.2018.2880814, 2169-3536 2018
IEEE. Translations and content mining are permitted for academic research only. Personal use is also
permitted, but republication/redistribution requires IEEE permission.
[13] Soudeh Babaeizadeh and Rohanin Ahmad. Constrained Artificial Bee Colony Algorithm for Optimization
Problems. AIP Conference Proceedings 1750, 020008 (2016); https://doi.org/10.1063/1.4954521 Published
Online: 21 June 2016.
[14] Soudeh Babaeizadeh and Rohanin Ahmad. Enhanced Constrained Artificial Bee Colony Algorithm for
Optimization Problems, The International Arab Journal of Information Technology, Vol. 14, No. 2, 2017.
[15] Nadezda Stanarevic, Milan Tuba, and Nebojsa Bacanin. Modified artificial bee colony algorithm for
constrained problems optimization. An International journal of mathematical models and methods in applied
science, issue 3, Volume 5, 2011.
[16] Jouni Lampinen, Ivan Zelinka. Mixed Variable non- linear optimization by differential evolution. Corpus ID:
11285973
[17] Md. Abul Kalam Azad, Edite MGP, Fernandes Ana MAC Rocha. Non-linear continuous global optimization
by modified differential evolution. Department of production and system school of Engineering, University of
Minho, 1710-057 Braga, Portugal.10.1063/1.3498653,2010.
[18] Musrrat Ali, Millie Pant, Ajith Abraham. Simplex differential evolution. Department of paper technology,
Indian institute of technology Roorkee Saharanpur capus, Saharanpur 247001 India vol 6, No.5, 2009.
[19] Zhang Xiao Fei, Guo Xiang Fulb, Yuan Li Hua. Reactive power optimization of power system based on
niching differential evolution algorithm. International symposium on computers and informatics (ISCI),2015.
[20] Chiha ibtissem, Liouqne Hend. In based cost differential evolution algorithm for multi objective optimization
problem. Vol 05. Electrica; Engineering department ENIM. Monastir, Tunisia. Asian general of Applied
sciences (ISSN 2321-0893),2017.

1018
Vol. 72 No. 1 (2023)
http://philstat.org.ph
 Mathematical Statistician and Engineering Applications
ISSN: 2094-0343
2326-9865

[21] Jeerayut Wetweerapong and Pikul Puphasak 2020: In an improved DEA with the restart technique to solve
system on non-linear equations. An International general of optimization and control theories and applications.
ISSN: 2146-0957, vol.10, No.1, PP.118-136 2020.
[22] Saeed nezhadhosein, Aghile Hejdari, Raza Ghanbari. Integrating DEA with modified Hybrid GA for solving
non-linear optimal control problems. International journal of mathematical science and informatics. Vol 12,
No.1 2017, PP (47-67)
[23] JR Koza, D.Andre, F. H. Bennett and M.Keane. Genetic algorithm Programming (3) of Darwinian Invention
and Problem Solving.
[24] David E. Genetic algorithms in search, optimization, and machine learning. Addison-Wesley Longman, USA;
1989.
[25] Dr. Amanpree Sings,Shivani Sanan, Amit Kumar. Solving Unconstrained Rosenbrock Optimization Problem
Using Genetic Algorithms. ISSN NO : 1006-6748, ISSN NO : 1006-6748, http://www.gjstx-e.cn,2021.
[26] D. Karaboga. An Idea Based On Honey Bee Swarm for Numerical Optimization, Technical Report-TR06,
Erciyes University, Engineering Faculty, Computer Engineering Department, 2005.
[27] B. Basturk, D. Karaboga. An Artificial Bee Colony (ABC) Algorithm for Numeric Function Optimization,
IEEE Swarm Intelligence Symposium 2006, May 12- 14, 2006, Indianapolis, Indiana, USA.
[28] J. Ilonen, J. K. Kamarainen and J. Lampinen. Differential evolution training algorithm for feed-forward neural
networks, Neural Processing Letters 17 (2003), no. 1, 93-105.

1019
Vol. 72 No. 1 (2023)
http://philstat.org.ph