You are on page 1of 8

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

net/publication/353197675

A New Hybrid Method for Global Optimization Based on the Bird Mating
Optimizer and the Differential Evolution

Conference Paper · February 2021


DOI: 10.1109/IEC52205.2021.9476147

CITATIONS READS

2 16

5 authors, including:

Haval Sadeeq Adnan Mohsin Abdulazeez


Duhok Polytechnic University Duhok Polytechnic University
8 PUBLICATIONS   38 CITATIONS    183 PUBLICATIONS   1,784 CITATIONS   

SEE PROFILE SEE PROFILE

Najdavan Abduljawad Kako Dilovan Zebari


Duhok Polytechnic University Universiti Teknologi Malaysia
7 PUBLICATIONS   28 CITATIONS    34 PUBLICATIONS   552 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Image Processing and Security View project

Different Model for Hand Gesture Recognition with a Novel Line Feature Extraction View project

All content following this page was uploaded by Diyar Zeebaree on 10 August 2021.

The user has requested enhancement of the downloaded file.


7th International Engineering Conference Research &Innovation amid Global Pandemic (IEC2021) Erbil, Iraq

A New Hybrid Method for Global Optimization Based on


the Bird Mating Optimizer and the Differential Evolution
2021 7th International Engineering Conference “Research & Innovation amid Global Pandemic" (IEC) | 978-1-6654-1283-4/20/$31.00 ©2021 IEEE | DOI: 10.1109/IEC52205.2021.9476147

Haval T. Sadeeq Adnan M. Abdulazeez Najdavan A. Kako


Information Technology Dept. Presidency of Duhok Information Technology Dept.
Duhok Polytechnic University, Polytechnic University, Duhok, Duhok Polytechnic University, Duhok,
Duhok, Kurdistan Region, Iraq Kurdistan Region, Iraq Kurdistan Region, Iraq
haval.tariq@dpu.edu.krd adnan.mohsin@dpu.edu.krd najdavan.kako@dpu.edu.krd

Dilovan A. Zebari Diyar Q. Zeebaree


Center of Scientific Research Director of Research Center of
and Development Nawroz University, Duhok Polytechnic University,
Duhok, Kurdistan Region, Iraq, Duhok, Kurdistan Region, Iraq
dilovan.majeed@nawroz.edu.krd dqszeebaree@dpu.edu.krd

Abstract—Nature-inspired algorithms are often used by several characteristics and concepts of biological systems. The
diverse areas of engineering and science due to their easiness and optimization process goes through crossover, mutation and then
versatility. Because metaheuristics operate by structurally the fittest gen will be selected [5]. Particle Swarm Optimization
changing and improving an established problem, they can often be (PSO) technique that is depends on the swarm conduct of both
extended to any optimization issues. The recent creation of meta- fish as well as birds, the system which is described as system of
heuristic algorithms has rendered them effective tools for solving
NP problems. This paper presents a hybrid meta-heuristic method
multi-agent may include features of this intelligence of swarm
based on the Differential Evolution and Bird Mating Optimizer group [6], Firefly Algorithm (FA) that is depends on the
techniques to solve problems of global optimization. Bird Mating features of flashing ideal conduct of fireflies in the tropic
Optimizer is a novel method and is inspired by mating behavior of regions [7][8] whereas another algorithm depending on the
birds. Bird Mating Optimizer has some drawbacks such as echolocation conduct of micro bats called a Bat Algorithm
producing poor results, trapping into local optima and slow (BA) [9]. Also, there are some recent nature inspired
convergence speed. Therefore, to conquer these insufficient it is algorithms; for example, Harris Hawks Optimizer (HHO) is an
hybridized with Differential Evolution approach. Differential optimization algorithm inspired by the conducts of cooperative
Evolution technique is utilized to retain a preferable balance as well as chasing patterns of predacious birds [10], Another
between both searches local and global. The performance and
effectiveness of new Differential Evolution and Bird Mating
algorithm which is simulates the conduct of slaps swarming
Optimizer algorithm is tested and evaluated on 15 different during transportation when foraging in oceans called the Salp
functions of benchmark. The results of the experiment have shown Swarm Algorithm (SSA) [11], also the technique of Beetle
the proposed technique possesses excellent performance in Antennae Search (BAS) which is inspired by the behavior of
convergence speed, stability, and robustness, as compared to the searching of long-horn bug [12], Finally, Black Widow
well-known algorithms. It is proved that the Differential Evolution Optimization (BWO) is an optimization technique that is
and Bird Mating Optimizer algorithm is very effective and inspired by the single conduct of black widow spiders mating
superior to solve problems of global optimization. Experimental [13].
results indicate that the proposed hybrid Differential Evolution Evolutionary algorithms look for the global optimal solution in
and Bird Mating Optimizer method is superior to previous
existing state-of-the-art metaheuristic algorithms.
a search space by arbitrarily generating one or more solutions
for a given problem. This set of solutions is considered the set
Keywords— Bird Mating Optimizer, Global Optimization, of nominee solutions. A selection of applicants to choose for
Hybrid Bird Mating Optimizer, Meta-heuristic Algorithms, the job is often improved iteratively until the fulfillment of a
Unconstrained Optimization Problems. terminating condition. The change may be viewed as
identifying a way to arrive specifically at the global maximum
rather than arbitrarily seek to achieve it. The mechanism of an
I. INTRODUCTION evolutionary algorithm offers many merits and benefits such as
dilemma independence, derivation independence, and
Modern optimization algorithms which typically based on depending on local optima [19].
swarm intelligence, are often nature-inspired [1][2]. There are In general, there are different optimization problems which
different manners for inspiration and accordingly techniques could be considered as constrained or unconstrained problems.
can be classified to several various types. Whilst almost all It is worth mentioning, that a constrained version of the present
these techniques have propensity to utilize some special and work has been proposed by the authors [14]. In this work, the
features for updating the objective function [3][4]. For example, algorithm will be validated and tested for problems of
genetic algorithm (GA) simulates Darwinian evolution unconstrained optimization utilizing fifteen benchmark

978-1-6654-1283-4/21/$31.00 ©2021 IEEE 54

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on July 14,2021 at 14:45:18 UTC from IEEE Xplore. Restrictions apply.
7th International Engineering Conference Research &Innovation amid Global Pandemic (IEC2021) Erbil, Iraq

functions and the result will be compared among some robust And when r1  mcf
and well-known algorithms. Askarzadeh [16] introduced a
BMO technique which is a nature-inspired meta-heuristic while xb (c) = l (c) − r2 *(l (c) − u (c)).
this technique still suffers from several drawbacks. For this (2)
reason and to obtain more precise evaluation results it is
hybridized with DE. To test the robustness and effectively of
proposed BMO-DE; the algorithm is implemented on some A random number is represented by c between 1 and 𝑛, the
functions of benchmark and the evaluation results compared to brood resultant indicated by 𝑥𝑏, the weight of time-varying to
other algorithms [25][26]. determine the significance of female that is interested by is
On the surface, an estimated algorithm (like a genetic represented by w . Where r is equal to vector of 1 d whose
algorithm) does not promise that the optimal solution will be each of their element randomly constructed in the rage of [1, 0]
discovered, but empirically this have also been shown to
influences the elements of the ( xi − x), n represents the
provide a nearly optimal solution within a fair period. There are dimension of problem. The factor of control mutation which
approximations that fall into two categories: intuitive
approximations and metaheuristics. There are several heuristics varies between 0 and 1 is indicated by mcf , random numbers
accessible in the literature, but generic heuristics do not extend are indicated by ri which are between 0 and 1 also. Finally, the
to any situation, with each problem getting its own heuristics.
upper bounds of the element are represented by
u, while the
Heuristic algorithms may be expensive to develop and might
not be ideal for a wide range of optimization problems [21]. In lower one is represented by l [14].
comparison, metaheuristics rely a lot more on feature Polygyny types which indicate that mating between one
optimization while provable items are more successful in a male of the bird with two females or several females. The major
more precise way. A high-level learning framework enables the feature of extra-pair intercourse is to obtain preferable genes for
creation of a higher-level conceptual model and adapts its brood. For more simplicity in the BMO technique, one brood
design for each cycle to better perform the current mission. only could be the outcome from the mating from genes which
Examples of these algorithms include simulated annealing, are a group of genes of females. The brood resultant can be
great deluge [21], tabu search [22], genetic algorithm (GA), performed using “(3)”. [16], [14].
[23], harmony search, scatter search, ant colony optimization
ni
xb = x + w *  rj * ( x j − x ) .
and particle swarm optimization [24][27].
The BMO-DE is motivated by the breeding tactics of male birds (3)
j =1
that feel it more advantageous to remain with one female to help
incubate a new offspring than to find new partners. The strength While when r1  mcf ( after that “(2)” is utilized). If ni
of the BMO-DE lies in its ability to have a strong compromise
indicates the number of birds that are interested, as well as xj
between discovery and extraction, i.e., the high exploration
areas and low exploration areas. The BMO-DE overcomes the represents the jth interesting bird. It is worth mentioning,
downside of lacking demographic diversity and being stuck in that the same computational equation of monogamous is
local optima that is inherent in other methodologies by using implemented to the promiscuous types.
five updating techniques to travel across the quest space. In parthenogenesis types, each female is raising the brood
without getting help form the male. In this species, and by the
II. BIRD MATING OPTIMIZER
way of modifying in probabilistically of her genes, females try
Bird Mating Optimizer (BMO) is a technique which is to pass-through on preferable genes to her brood. The outcome
based on the population algorithms. The population in the BMO
algorithm is suggested to society and for a specific problem of each bird of a brood based on the “(4)” and “(5)”: if r > 1

each society member can be considered as a feasible solution mcfp .


and is named a bird. Usually, genes in high quality are possess
by females which can be divided into different sets, namely,
promiscuous, monogamy, and polygyny [16].
xb (i) = x(i) +  *(r2 − r3 )* x(i).
In monogamy types, to select one female for mating, the (4)
quality of females will be assessed by each individual male by Else
using a probabilistic approach. The females who possess genes
in high quality have more chances to be chosen. xb (i) = x(i).
Computationally manner, suppose x is a monogamous bird
(5)
x , the following If
 is the size step and mcfp indicates the factor of
which wants to make mating with a female
control for parthenogenetic mutation. In the BMO technique, a
Equations give the brood resultant [16].
similar computational equation of polygyny is implemented to
xb = x + w * r * ( xi − x ) . the polyandry types.
(1)

55

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on July 14,2021 at 14:45:18 UTC from IEEE Xplore. Restrictions apply.
7th International Engineering Conference Research &Innovation amid Global Pandemic (IEC2021) Erbil, Iraq

III. DIFFERENTIAL EVOLUTION (DE) random number. Thus, the jth component of vi is modified
Differential Evolution (DE) technique is a stochastic as shown in “(8)” [4]:
approach which is depended on the population which is
presented by Storn and Price [17]. This approach represents the
design of vectors or parameters in a d-dimensional via search
space. Preserving candidate solutions of a population is the first
u tj+,i1 = v j ,i if ri  Cr .
(8)
step of the optimization process starts in DE. Then, based on
the combining of existing candidate solutions a new one could While no changes will be happening on it, randomly this
be generated depending on its objective function. The final step technique can determine whether any exchanges could happen
is preserving the candidate solution which has an optimal value between each component and donor vector.
of fitness at hand [20].
Dealing with a mutation in DE could be based on randomly
selecting various population vectors for two different vectors, C. Selection
after that, utilized for the disorder the presenting vector. The process of selection is related to the fittest genes
Further, such disorder is implemented in each population selection in the DE technique as well as in the Genetic
Algorithm (GA). Literature focuses on Cr , F and n selection
vector, thus, it is considered as a very effective approach.
Moreover, it could be mentioned that the process of crossover
in the DE approach is a vector-based also. Here, in the case of also amending of “(7)”. Generally, different methods of “(7)”
having d vectors in the d-dimensional of search space as well formulation can be utilized to provide vectors of mutation, this
as generation of first random of n solution vectors, then helps in leading to various algorithms with the calling of the
selecting xi if i = 1, 2,3,..., n for each solution xi at any convention. Where the scheme of mutation indicated by x , the
y
generation t , based on that, the conventional notation can be number of the variation vector is represented by , and z is
chosen as the following ” (6)” . considered as a scheme of crossover (exponential or binomial).
“(7)” presents the basic of DE/Rand/1/Bin methods [4], [20].

xit = ( x1,t i , x2,t i ,..., xdt ,i ), IV. PROPOSED METHOD


(6)
This study introduced a new hybrid technique based on
The DE approach consists of some major processes which meta-heuristic algorithms. The proposed study exploits the
are mutation process, crossover process, and selection process. original methods of Bird Mating Optimizer (BMO) and
Differential Evolution (DE) techniques to propose a new
method. A hybrid method between BMO and DE has been
A. Mutation presented to outcome Unconstrained optimization problems.
In this step, three featured vectors are selected randomly BMO suffers from many drawbacks like other meta-
heuristic algorithms, such as the convergence speed, trapping
xp, xqand xr at t for each vector xi at any producing t , to into local optima and the poor solution quality. Therefore, and
produce a vector which is named as the donor vector the to overcome these insufficiencies, BMO is hybridized with DE
following “ (7)” could be used [4]. in this paper. Basically, to maintain a preferable balance
between both searches local and global, as intense global search
has utilized the DE technique in this study. Thus, the proposed
vit +1 = xtp + F ( xqt − xrt ), method does not be trapped in local optimum. The optimization
(7) process of the proposed BMO-DE starts by initializing
Where F indicates weight factor of real differential and parameter settings and population size, then the maximum
 [0, 2] , the which is named a donor vector vi is produced number of generations, polygynous, percentage of
monogamous, polyandrous birds, and promiscuous will be
from the perturbation process F ( xq − xr ) to the vector xp , initialized, after that, the objective (fitness) function will be
and the size of population n must be equal or more than number calculated, and the birds will be sorted out based on the value
of their objective function. Different birds, namely,
four.
polyandrous, Monogamous, and polygynous will be
B. Crossover determined, then based on chaotic sequence, all species will be
generated and the worst one will be deleted. Then the role of
The probability of crossover cr  [0,1] can control the the DE approach will appear as intensive global search to select
procedure of crossover. The procedure of crossover is executed elite birds.
through two methods: binomial and exponential. This In the technical meaning, in the DE technique, both
procedure is implemented though the scheme of binomial on crossover and mutation processes can be implemented in
each of the d components by r [0;1] which represents a
i several manners. Thus, a particular notation was utilized to
describe the variety of these strategies or approaches. For
instance, “DE/rand/1/bin" indicates that the major vector is

56

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on July 14,2021 at 14:45:18 UTC from IEEE Xplore. Restrictions apply.
7th International Engineering Conference Research &Innovation amid Global Pandemic (IEC2021) Erbil, Iraq

chosen randomly, where the difference of one vector could be carefully depending on the achieved ideal evaluation results. To
added to it. The number of amended parameters in the vector of be adequate, the number of elite mates who are interested in for
the mutation process follows the binomial allocation. Finally, birds polyandrous as well as polygynous is set to three. Among
the objective function will be updated, and if the termination 10 of monogamous birds, the interested monogamous birds
criteria are met, then the optimization process will be have been chosen with genes which are most promised to
terminated. Fig. 1 shows the pseudo code of the proposed depend on the original method of BMO. The value of mcf
method utilizing the DE/ Rand/ 1/ Bin scheme. which indicates the factor of the control of mutation is set to 0.9
1: Set and initialize the size of the society (population size) whereas the
 which represents the step size is set to 0.001.
2: Initialize maximum no. of generations, percentage of monogamous,
polygynous, promiscuous, and polyandrous birds and other algorithm The entire process may be affected by the not tunning the mcf
parameters. parameter carefully which leads to have a bad impact on the
3: Do evaluation performance. The time-varying weight factor
indicated by w is set between the range of 2.5 to 0.25, the
4: Calculate the objective function
5: Sort the birds according to their objective function value
6: Determine monogamous, polygynous, and polyandrous birds factor of parthenogenetic mutation control is represented by
7: Based on chaotic sequence, generate promiscuous birds and delete the
worst birds
mcfp is set between the rage of 0.1 to 0.9. These values are
8: Calculate the objective function for the promiscuous birds set to present a good balance between both searches global and
9: For i=1 to maximum no. of promiscuous birds local. Hence, 100 is set as the maximum number of generations.
10: Select and choose the elite bird using DE/ Rand/ 1/ Bin scheme
11: The brood is produced based on equations 1 and 2
For DE, the following parameters are employed: the mutation
12: Next i strategies: \DE/rand/1", the factor of differential amplification
13: For i=1 to maximum no. of polygynous birds F = 0.52, constant probability of crossover CR = 0.91.
14: Select and choose the elite bird using DE/ Rand/ 1/ Bin scheme
15: The brood is produced based on equations 3 and 4 B. Stopping Criteria in BMO-DE
16: Next i
17: For i=1 to maximum no. of polygynous birds
The number of iterations for prediction is nearly considered
18: Select and choose the elite bird using DE/ Rand/ 1/ Bin scheme as a standard termination procedure in whole techniques of
19: The brood is produced based on equations 3 and 4 swarm intelligence. Thus, the optimal solution may be missed
20: Next i by the specific algorithm. In this study, the largest number of
21: For i=1 to maximum no. of promiscuous birds iterations placed as 100 iterations for the termination procedure
22: Select and choose the elite bird using DE/ Rand/ 1/ Bin scheme and with a given error rate (tolerance) of 0.00001. This to
23: The brood is produced based on equations 1 and 2
20: Next i ensure that one of the previous stop conditions will discover the
21: For i=1 to maximum no. of parthenogenetic birds optimal solution.
22: The brood is produced based on equations 5 and 6
23: Next i C. Functions of Benchmark
24: Calculate the objective function The performance evaluation of the proposed hybrid BMO-
25: Perform the replacement and update the parameters until the
termination criteria is met DE method has been assessed based on employing 15 standard
functions of the benchmark with different properties, this
evaluation has been done from various perspectives.
Fig. 1. pseudo code of the proposed BMO-DE algorithm Benchmark functions could be classified into different three
sets: unimodal results as shown in Table I, the result of
V. EXPERIMENTAL RESULT AND DISCUSSION multimodal as shown in Table II, and finally the results of
In this study, the experimental evaluation is performed in composite functions as shown in Table III. As their names
MATLAB R2020a, and carried out on a LENOVO Laptop with imply, the functions of the unimodal test consist of only one
the configuration of Intel® Core™ i7-8550U CPU @ 2.00 GHz optimal solution. In the other side, multi-modal test functions
with 8.00 GB RAM on Windows 10 of used operating system. usually consist of two or more optimum, so, it is more tough
than unimodal functions. The optimum points in multimodal
functions are called local optima except the main one which is
A. BMO-DE Parameter Settings called global optima. A robust algorithm shall not be trapped in
In metaheuristic algorithms, the set of parameters must be any of these local optima points to method and reach the global
tuned carefully as it affects the performance evaluation of the optimum. Thus, discovering, and local optimal avert of
method. Wherefore, it is better that the parameter settings of the technique can be evaluated based on the functions of test of the
proposed BMO-DE will be studied more soon. The parameter multi-modal.
settings in this work are as the following: It is worth mentioning, all test functions of both unimodal
The size of both population and the society has placed to functions and multi-modal functions are equivalent to 0 as
200. The initialization of whole birds has been done in a random minimally. The last fourth functions used for testing are
way via the search space. The number of both polyandrous as composite functions that are combined extremely, shifted,
well as parthenogenetic is set at 10, the number of promiscuous rotated, and biased version of unimodal test functions and
has set at 20, the number of polygynous has set at 60, and 100 multi-modal test functions [18]. These functions simulate the
has been set as the number of monogamous birds. Further, the real spaces of search by supplying a huge number of local
ratio of each type from the population has been selected optimum and various shapes of the space of search. A good
method should have an excellent balance between

57

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on July 14,2021 at 14:45:18 UTC from IEEE Xplore. Restrictions apply.
7th International Engineering Conference Research &Innovation amid Global Pandemic (IEC2021) Erbil, Iraq

diversification and intensification to reach the global optima of 𝑓11(𝑥) 30, 200 [-50, 50] 0
such functions of test. = 0.1 {𝑠𝑖𝑛2 (3𝜋𝑥1 )
𝑛
TABLE I. UNIMODAL BENCHMARK FUNCTIONS. +∑ (𝑥𝑖
𝑖=1
2 [1
Functions Dims Range Fmin − 1) + 𝑠𝑖𝑛2 (3𝜋𝑥1 + 1)] + (𝑥𝑛
𝑛
30, 200 [-100, 100] 0 − 1) 1 + 𝑠𝑖𝑛2 (2𝜋𝑥𝑛 )]
2

𝑓1(𝑥) = ∑ 𝑥𝑖2
𝑖=1
𝑛 𝑛
30, 200 [-10, 10] 0 TABLE III. BENCHMARK FUNCTIONS COMPOSITE
𝑓2(𝑥) = ∑ 𝑥𝑖2 |𝑥𝑖 | ∏ |𝑥𝑖 |
𝑖=1 𝑖=1/2
𝑖 30, 200 [-100, 100] 0 Functions Dims Range Fmin
𝑛
𝑓3(𝑥) = ∑ 𝑥𝑖2 (∑ 𝑥𝑗 ) 𝑓12(𝑥) = ∑11
𝑥1 (𝑏𝑖2 +𝑏𝑥2 )
] 4 [-5, 5] 0.0003
𝑖=1 𝑖=1[𝑎𝑖 −
𝑏𝑖2 +𝑏𝑥3 +𝑥4
𝑗−1
𝑓4(𝑥) = max 𝑖 {|𝑥𝑖 |, 1 ≤ 1 ≤ 𝑛} 30, 200 [-100, 100] 0
1 2 [-5, 5] -1.031
𝑓13(𝑥) = 4𝑥12 − 2.1𝑥14 + 𝑥16 + 𝑥1 𝑥2
3
𝑛−1
30, 200 [-30, 30] 0 + 4𝑥22 + 4𝑥24
𝑓5(𝑥) = ∑ [100(𝑥𝑖+1 − 𝑥𝑖2
𝑖=1
5.1 5
𝑓12(𝑥) =(𝑥2 − 2 𝑥12 + 𝑥1 − 6)2 + 2 [-5, 5] 0.398
4𝜋 𝜋
+ (𝑥𝑖 − 1)2 ] 1
𝑛
30, 200 [-100, 100] 0 10 (1 − ) 𝑐𝑜𝑠𝑥1 + 10
8𝜋
𝑓6(𝑥) = ∑ ([𝑥𝑖 + 0.5)2 ]
𝑖=1 𝑓12(𝑥) =[1+(𝑥1 + 𝑥2 + 1)2 (19 − 2 [-2, 2] 3
𝑛
30, 200 [-128, 128] 0 14𝑥1 + 3𝑥12 − 14𝑥2 + 6𝑥1 𝑥2 + 3𝑥22 )] ∗
𝑓7(𝑥) = ∑ 𝑖𝑥𝑖4 + 𝑟𝑎𝑛𝑑𝑜𝑚[0,1) [30 + 2𝑥1 − 3𝑥2 )2 ∗ (18 − 32𝑥1 +
𝑖=1
12𝑥12 + 48𝑥2 − 36𝑥1 𝑥2 + 27𝑥22 )]
TABLE II. MULTIMODAL BENCHMARK FUNCTIONS.
Functions Dims Range Fmin
To verify and validate the BMO-DE results, three well-
known and robust techniques are employed: the preferable
𝑛
30, 200 [-5.12, 0
𝑓8(𝑥) = ∑ [𝑥𝑖2 − 10 cos(2𝜋𝑥𝑖 )
5.12]
evolutionary technique is a GA while PSO from swarm
intelligence techniques is considered as the ideal one.
𝑖=1
+ 10]
𝑛 30, 200 [-32, 32] 0 Additionally, the BMO-DE technique is compared with one of
1
𝑓9(𝑥) = −20exp⁡(−0.2√ ∑ 𝑥𝑖2 ) the modern and robust techniques which is Bat Algorithm (BA)
𝑛
𝑖=1 [19] Table IV, Table V, and Table VI.
𝑓10(𝑥) 30, 200 [-600, 600] 0
To obtain the performance of quantitative evaluation
1
𝑛
𝑛 𝑥𝑖
results, each of the used technique has been performed on the
=
4000
∑ 𝑥𝑖2 − ∏ cos⁡( ) functions of test 30 times independently. The average of the
𝑖=1 √𝑖
𝑖=1 minimum objective values is denoted by (AVE) and standard
+1
deviation (STD) of the best ideal solution until stopping criteria
are met is reported. The two criteria strategies indicate which
technique act as more stable technique during solving the test
functions.

TABLE IV. RESULTS OF PERFORMANCE EVALUATION OF UNIMODAL BENCHMARK FUNCTIONS (30-DIMENSIONAL)

GA PSO BA BMO-DE
F
AVE STD AVE STD AVE STD AVE STD
F1 0.118842 0.125606 2.70E_09 1.00E_09 0.773622 0.528134 2.68E_11 1.89E_10

F2 0.145224 0.053227 7.15E_05 2.26E_05 0.334583 3.816022 1.98E_05 5.37E_06

F3 0.13902 0.121161 4.71E_06 1.49E_06 0.115303 0.766036 5.07E_7 5.84E_8


F4 0.157951 0.862029 3.25E_07 1.02E_08 0.192185 0.890266 1.44E_07 1.92E_09

F5 0.714157 0.972711 0.123401 0.216251 0.334077 0.300037 0.356883 0.132446


F6 0.167918 0.868638 5.23E_07 2.74E_06 0.778849 0.67392 2.78E_8 1.11E_9
F7 0.010073 0.003263 0.001398 0.001269 0.137483 0.112671 0.003485 0.004143

58

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on July 14,2021 at 14:45:18 UTC from IEEE Xplore. Restrictions apply.
7th International Engineering Conference Research &Innovation amid Global Pandemic (IEC2021) Erbil, Iraq

TABLE V. PERFORMANCE EVALUATION RESULTS OF MULTIMODAL BENCHMARK FUNCTIONS (30-DIMENSIONAL)


GA PSO BA BMO-DE
F
AVE STD AVE STD AVE STD AVE STD

F8 0.659271 0.815751 0.278588 0.218991 1.233748 0.686447 8.86365E_04 8.65E_05

F9 0.956111 0.807701 1.11E_09 2.39E_11 0.129359 0.043251 4.83437E_11 1.45E_12

F10 0.487809 0.217782 0.273674 0.204348 1.451575 0.570309 0.027515 0.008993

F11 0.110769 0.002152 9.42E_09 2.31E_10 0.395977 0.993325 0.012654 9.47E_11

TABLE VI. PERFORMANCE EVALUATION RESULTS OF COMPOSITE BENCHMARK FUNCTIONS

GA PSO BA BMO-DE
F
AVE STD AVE STD AVE STD AVE STD

F12 114.6139 26.96248 120 131.6561 182.476 117.0248 0.891365 0.991354

F13 95.46331 7.163383 162.9144 119.2351 487.2021 161.4107 58.134262 30.392763

F14 325.4427 51.66827 363.2361 151.3109 588.1938 137.7861 201.253146 76.301020

F15 466.3074 29.56841 450.0688 157.8496 756.9757 160.097 310.014151 11.030561

Based on the obtained evaluation results of the proposed as the largest number of iterations (generations) in the first
BMO-DE method using various benchmark test functions stage. Further, computing the objective function as well as
presented in Table I, Table II, and Table III. Further, comparing sorting out whole birds depended on the value of their objective
obtained results with some well-known and robust optimization function utilizing the BMO technique. After that, based on
techniques presented in Table IV, Table V, and Table VI. It was utilizing DE technique the best group of elite birds will be
noticed that the present work outperforms other studies from chosen. In the experimental evaluation, various benchmark test
state-of-the-art almost in all test functions. DE technique is the functions have been utilized which are unimodal, multimodal,
main reason behind achieving the ideal results of the proposed and composite functions. This step has been done to ensure the
method. The DE technique possesses the exploitation or robustness and efficacy of the proposed technique. After that,
intensification property, which uses the data and information of this study presents a comparison between proposed technique
the existing solutions based on focusing on a local scale for with several literature techniques in term of both STD and
searching. Moreover, another important property can be found average of the best optimal solution until stopping criteria are
in the DE technique, which is exploration or diversification, met. The experimental evaluation results depicted that the
depending on that the DE technique can produce diverse or proposed BMO-DE method is approximately reliable and
different solutions for exploring search space at the global level. effective in finding a global optimal solution accurately.
Based on that, the proposed BMO-DE obtained the global Eventually, it can be acknowledged that the proposed hybrid
optimality since a good balance and incorporation which are BMO-DE technique is a promise and efficient contribution to
provided between diversification and intensification. overcoming these types of optimization problems.

VI. CONCLUSION REFERENCES


This study presents a new hybrid meta-heuristic method
named the BMO-DE technique. Based on the proposed [1] Y. XS (2010) Nature-Inspired Metaheuristic Algorithms, vol 2.
technique could overcome global optimization problems as Nature-Inspired Metaheuristic Algorithms, vol 2, Second Edition
edn. Luniver Press, United Kingdom
well as more specifically unconstrained ones. Initialization is [2] D. Q. Zeebaree, A. M. Abdulazeez, D. A. Zebari, H. Haron, and N.
considered as a first step of the optimization process for the A. Haza, "Multi-Level Fusion in Ultrasound for Cancer Detection
proposed technique. Population size has been initialized as well Based on Uniform LBP Features". Computers,Materials &
Continua, Vol.66, No.3, pp. 3363- 3382, 202.

[3] A.S. Eesa, Z. Orman, & A.M. Brifcani, (2015). A novel feature-
selection approach based on the cuttlefish optimization algorithm
for intrusion detection systems. Expert Syst. Appl., 42, 2670-2679.
[4] D. Q. Zeebaree, H. Haron, A. M. Abdulazeez, and S. R. Zeebaree,
‘Combination of K-means clustering with Genetic Algorithm: A

59

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on July 14,2021 at 14:45:18 UTC from IEEE Xplore. Restrictions apply.
7th International Engineering Conference Research &Innovation amid Global Pandemic (IEC2021) Erbil, Iraq

review’, Int. J. Appl. Eng. Res., vol. 12, no. 24, pp. 14238–14245, Engineering (ICOASE), Duhok, 2018, pp. 145-150, doi:
2017. 10.1109/ICOASE.2018.8548836.
[5] H. Sadeeq,“A Modified Flower Pollination Algorithm for [16] A. Askarzadeh (2014) Bird mating optimizer: An optimization
Engineering Design Optimization Problems”. Polytechnic Journal. algorithm inspired by bird mating strategies. Communications in
Vol 7, No 1. (2017). Nonlinear Science and Numerical Simulation 19 (4):1213-1228.
[6] D. Q. Zeebaree, H. Haron, A. M. Abdulazeez, and D. A. Zebari, doi: 10.1016/j.cnsns.2013.08.027
"Trainable Model Based on New Uniform LBP Feature to Identify [17] R. Storn KP (1997) Differential Evolution – A Simple and Efficient
the Risk of the Breast Cancer," in 2019 International Conference on Heuristic for global Optimization over Continuous Spaces. Journal
Advanced Science and Engineering (ICOASE), 2019: IEEE, pp. of Global Optimization 11 (4):341–359.
106-111. doi:10.1023/A:1008202821328
[7] R. Zebari, A. Abdulazeez, D. Zeebaree, D. Zebari, and J. Saeed, "A [18] J. Liang, B. Qu, & P.N. Suganthan, (2014). Problem Definitions
Comprehensive Review of Dimensionality Reduction Techniques and Evaluation Criteria for the CEC 2014 Special Session and
for Feature Selection and Feature Extraction," Journal of Applied Competition on Single Objective Real-Parameter Numerical
Science and Technology Trends, vol. 1, no. 2, pp. 56-70, 2020. Optimization.
[8] H. Sadeeq and A. M. Abdulazeez, "Hardware Implementation of [19] S. Mirjalili, The ant lion optimizer, Adv. Eng. Software, 83 (2015)
Firefly Optimization Algorithm Using FPGAs," 2018 International 80-98.
Conference on Advanced Science and Engineering (ICOASE), [20] Qin, A. K., Huang, V. L., & Suganthan, P. N. (2008). Differential
Duhok, 2018, pp. 30-35, doi: 10.1109/ICOASE.2018.8548822. evolution algorithm with strategy adaptation for global numerical
[9] D.A. Zebari, D.Q. Zeebaree, J.N. Saeed, N.A. Zebari, N& A.Z. optimization. IEEE transactions on Evolutionary
Adel,. (2020). Image steganography based on swarm intelligence Computation, 13(2), 398-417.
algorithms: A survey. people, 7(8), 9. [21] M.K. Dhadwal, S.N. Jung, C.J. Kim, Advanced particle
[10] A.A. Heidari, et al. “Harris Hawks Optimization: Algorithm and swarm assisted genetic algorithm for constrained optimization
Applications.” Future Generation Computer Systems, Elsevier BV, problems, Comput. Optim. Appl. 58 (2014) 781–806.
Feb. 2019, doi:10.1016/j.future.2019.02.028. [22] E. Taha Yassen, M. Ayob, M.Z. Ahmad Nazri, N.R. Sabar,
[11] S.M. Mirjalili, et al. “Salp Swarm Algorithm: A bio-inspired Metaharmony search algorithm for the vehicle routing problem with
optimizer for engineering design problems.” Adv. Eng. Softw. 114 time windows, Inf. Sci. (Ny). 325 (2015) 140–158.
(2017): 163-191. [23] K. Deb, N. Padhye, Enhancing performance of particle swarm
[12] D. Q. Zeebaree, H. Haron, A. M. Abdulazeez, and D. A. Zebari, optimization through an algorithmic link with genetic algorithms,
"Machine learning and Region Growing for Breast Cancer Comput. Optim. Appl. 57 (2014) 761–794.
Segmentation," in 2019 International Conference on Advanced [24] A. Arram, M. Ayob, G. Kendall, & A. Sulaiman, (2020). Bird
Science and Engineering (ICOASE), 2019: IEEE, pp. 88-93. mating optimizer for combinatorial optimization problems. IEEE
[13] V. Hayyolalam, A. Asghar Pourhaji Kazem, “Black Widow Access, 8, 96845-96858.
Optimization Algorithm: A novel meta-heuristic approach for [25] M. Younis, S.Y. Ameen, S.B. Sadkhan, “evaluation of using genetic
solving engineering optimization problems”. Engineering algorithm in cryptoanalysis”, Journal of University of Babylon,
Applications of Artificial Intelligence. 87 (2020), 2006, 12 (5), 982-989.
https://doi.org/10.1016/j.engappai.2019.103249. [26] S.B. Sadkhan, “A Proposed Genetic Based Method to Solve
[14] H. Sadeeq ., A. Abdulazeez ., N. Kako , A. Abrahim (2018) A Novel Frequency Assignment Problem for HF Band”, 2019 4th Scientific
Hybrid Bird Mating Optimizer with Differential Evolution for International Conference Najaf (SICN), 48-53.
Engineering Design Optimization Problems. In: Saeed F., Gazem [27] S.B. Sadkhan, F.H. Abdulraheem, “A proposed ANFIS Evaluator for
N., Patnaik S., Saed Balaid A., Mohammed F. (eds) Recent Trends RSA Cryptosystem used in Cloud Networking”, 2017 International
in Information and Communication Technology. IRICT 2017. Conference on Current Research in Computer Science and
Lecture Notes on Data Engineering and Communications Information Technology (ICCIT)..
Technologies, vol 5. Springer, Cham. https://doi.org/10.1007/978-
3-319-59427-9_55
[15] D. Q. Zeebaree, H. Haron and A. M. Abdulazeez, "Gene Selection
and Classification of Microarray Data Using Convolutional Neural
Network," 2018 International Conference on Advanced Science and

60

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on July 14,2021 at 14:45:18 UTC from IEEE Xplore. Restrictions apply.
View publication stats

You might also like