Received December 24, 2018, accepted January 7, 2019, date of publication January 17, 2019, date of current version

February 8, 2019.
Digital Object Identifier 10.1109/ACCESS.2019.2893662

Comprehensive Review of the Development

of the Harmony Search Algorithm and
Its Applications
ALA’A A. AL-OMOUSH , ABDULRAHMAN A. ALSEWARI , (Senior Member, IEEE),
HAMMOUDEH S. ALAMRI, AND KAMAL Z. ZAMLI , (Member, IEEE)
Faculty of Computer Systems and Software Engineering, Universiti Malaysia Pahang, Kuantan 26300, Malaysia
Corresponding author: Abdulrahman A. Alsewari ([email protected])
This research is partially funded by FRGS/1/2018/ICT05/UMP/02/1, RDU190102: A Novel Hybrid Kidney-inspired algorithm for Global
Optimization, and UMP PGRS160396: Hybridize Jaya Algorithm for Harmony Search Algorithm’s Parameters Selection.

ABSTRACT This paper presents a comprehensive overview of the development of the harmony search (HS)
algorithm and its applications. HS is a well-known human-based meta-heuristic algorithm that mimics the
process of creating a new harmony in music. This algorithm can be applied to different fields of research,
owing to its ability to balance between exploitation (i.e., searching around the known best) and exploration
(i.e., roaming the entire search space). Thus, numerous studies have been conducted to utilize HS in
real-world optimization problems, and many variants and hybrid algorithms of HS have been developed
to cope with different problems. In this paper, HS and its variants are reviewed from various aspects.
First, we describe the HS algorithm and present how its parameters affect algorithm performance. Second,
we describe HS classifications based on the well-known HS variants and hybrid algorithms, along with
their applications. Finally, a discussion conducted on the strengths and weaknesses of the HS algorithm and
the possibilities for its improvement. Focusing on related work from diverse fields (such as optimization,
engineering, computer science, biology, and medicine), this paper can foster interests on the application of
HS for multidisciplinary audiences.

INDEX TERMS Harmony search, evolutionary algorithms, meta-heuristics, Optimization, computational

intelligence.

I. INTRODUCTION a limited time [4]. The disadvantage of this method is the need
Optimization algorithms aim to find the fittest element from to create a specific solution for a specified problem. Hence,
groups of choices subjected to specific constraints [1], [2]. one algorithm cannot be used for all types of problems.
The use of the optimization algorithm to solve real-world Meta-heuristic algorithms are higher-level heuristic algo-
problems started in the early 1940s. The optimization algo- rithms that can cover a wider range of problems [5]. The
rithms have two main categories. The first category includes main functionality of meta-heuristic algorithms is obtained
exact algorithms, which prove their ability to find optimal by merging rules and randomness to simulate natural phe-
solution values in direct and typical models. However, real nomena, such as physical annealing in a simulated anneal-
world optimization problems are complicated, and thus, are ing (SA) algorithm [6], human intelligence in the harmony
impossible to solve using exact algorithms [3]. The second search (HS) algorithm [7], the biological evolutionary process
category includes heuristic algorithms, which are developed in differential evolutionary algorithm (DE) [8], and animal
to solve problems faster than classic methods by scarifying behavior in Tabu search [9].
optimally or completeness to find a near optimal value. The effectiveness of meta-heuristic algorithms relies on
Heuristic optimization technique contains a general design the utilization of explorative (diversification) and exploitative
that does not require specific conditions or mathematical (intensification) ranges through a search [10]. The exploita-
attributes of objective functions, such as gradient information tive process is achieved by utilizing the information gained
and differentiable attributes; thus, researchers rely more on to guide the search toward its goal. The explorative process is
heuristic algorithms to find the most optimal solutions within defined as the ability of an algorithm to investigate uncovered
2169-3536 2019 IEEE. Translations and content mining are permitted for academic research only.
VOLUME 7, 2019 Personal use is also permitted, but republication/redistribution requires IEEE permission. 14233
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

areas rapidly within large search sizes. Implementation per-

formance will improve if the balance between these two
features is achieved [11].
Some researchers categorize meta-heuristic algorithms
based on their method of searching, such as trajectory-based
or population-based algorithms [12]. The method used
in searching has a special effect on optimization algo-
rithms. For example, SA works through functional search
nodes or agents that shape the Brownian motion path
through its movement across certain targets. Meanwhile,
population-based algorithms are implemented through the
parallel use of various operators. Although both techniques
have achieved successful results in different works, some
studies suggest that population-based algorithm is more suit-
able for multi-objective optimization problems [13]. How- FIGURE 1. The number of HS publications per database for several
ever, combining these two techniques can produce good journals: ACM, Elsevier, Hindawi, IEE, Springer, Taylor & Francis, and other
journals as one group.
results [4].
Geem et al. [7] proposed a well-known population-based
meta-heuristic algorithm, known as harmony search algo-
rithm(HS). This algorithm mimics the process of creating
a new harmony in music to solve an optimization prob-
lem. HS has been proven to produce exceptional results
over a vast scope of optimization problems because of its
ability to handle different optimization problems in various
fields, such as university timetable [2], [14], [15], structural
design [3], [10], [16]–[18], water distribution [19]–[21], and
other fields of research. The main advantage of the HS
algorithm is that it is easy to code and apply to diverse
problems [22].
The ability of the HS algorithm to achieve a balance
between exploitative and explorative ranges is the reason for
its strength and success. In the HS process, the exploitative
range is mainly dominated by pitch adjustment rate (PAR) and
bandwidth (BW), whereas the explorative range is essentially
controlled by the HS memory-accepting rate (HMCR) [2].
This review aims to present an overall summary of related FIGURE 2. The number of publications between (1999 – 2018).
studies on the HS algorithm, including its characteristics,
variants, hybridizations, applications to different optimiza-
tion fields, and future challenges and potentials. In this work,
we classified HS algorithm research topics into four cate- hybrid HS and provides examples. Section 5 presents HS
gories: HS general structure, classifications, hybridizations, applications. Finally, Section 6 includes discussions and con-
and applications. cludes the paper.
In this article, we determined the number of HS publi-
cations from various publishers: ACM, Hindawi, Taylor & II. THE GENERAL STRUCTURE OF THE HS ALGORITHM
Francis, Elsevier, IEEE, SpringerLink, and others. Fig 1 HS is a well-known (EA). EA basically begins by creating
shows the number of published papers based on each database random values that are possible solutions to a specific prob-
when the HS algorithm is applied to various optimization lem. The fitness of each value is evaluated based on the
fields. Meanwhile, Fig 2 shows the growing interest in the evaluation function. A nomination process must be used in
utilization of the HS algorithm and its applicability to differ- every rotation within EA to create a better population. The
ent fields of research compared to other recent metaheuristics main objective of the nomination process is to deviate the
based on the Google Scholar database for the period from search toward fitter values, capture these values, and include
1999 until 2018. them in the next population [23].
The rest of this paper is organized as follows. Section 2 The HS algorithm is applied in different areas due to the
presents the general structure of HS. Section 3 discusses the following special advantages: HS requires lower mathemati-
effects of HS parameters and identifies well-known HS clas- cal operations than traditional heuristic algorithms, does not
sifications and several HS-based variants. Section 4 describes use premier amount or gradient search for decision variables,

14234 VOLUME 7, 2019

A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

and does not consider the derived values because stochastic

inspections are used [3].
The main similarities between the HS algorithm and
music harmony improvisation processes are as follows. First,
a musician searches for the most beautiful music harmony
(ideal condition) as specified by the standard for tunes, while
the HS algorithm searches for the global solution (ideal con-
dition) as specified by an objective function. Second, every
pitch of a musical instrument contributes to the level of music
harmony and beauty, which is similar to the objective function
value specified by the combination of values assigned to each
decision variable [3], [10].
HS involves processes that are analogous to those of other
meta-heuristic algorithms. For example, it saves the history
of visited vectors in harmony memory (HM), which is com-
parable with the memory of the Tabu search algorithm. HS is
capable of mutating the (HMCR) through the entire calcula-
tion process, similar to SA. HS considers all vectors, which
is similar to a genetic algorithm (GA). The main difference
between HS and GA is that the latter creates a new vector
based on two previous vectors (called parents). By contrast,
HS considers all previous vectors before creating a new har-
mony vector [10].
The HS algorithm process contains five main steps,
as illustrated in Fig 3.
Step 1: Initialize values of the HS parameters, such as
HMCR, BW, PAR, number of iterations (NI), and HM size
(HMS).
The optimization problem goal will be determined in this
step, either by using maximum or minimum objective func-
tion f (xi ), where xi will be the possible solution from N (N
∈ all the decision variables of xi ).
Step 2: HM values will be initialized, as xi within the upper
and lower boundary ranges, using the following equation:

xi = lowerbound + R1 × (upperbound − lowerbound)

FIGURE 3. HS flow chart.
.{R1 is random number (0 − 1)} (1)

Step 3: Improvisation of the new harmony is performed

using a combination of three major parameters: HMCR, PAR, III. HS CLASSIFICATIONS
and BW. The improvisation has two main steps, as shown in The following parameters affect the performance of HS:
Algorithm 1. First, two random values (a & b) will be created HMS, HMCR, BW, and PAR. Many researchers have tried
between (0 ∼ 1), and if (a > HMCR), then a new value xj will to enhance the performance of HS algorithm by selecting
0
be created using Eq. (1), as a new generated vector (xj ). the ideal combination of these variables [24]. In this section,
Second If the value of (a < HMCR), then a random value we will show the descriptions of these parameters from previ-
from HM (xi ) will be selected, and if the value of (b < PAR), ous studies, some suggested values, and how they affect HS.
then xi will be modified using Eq. (2). HMCR enables selecting one of the existing values in HM
0
xj = xnew,j ± bw ∗ rand (2) (xi,j ) or creating a new value xjnew through the improvisation
process; that is, the larger the value of HMCR, the larger
Step 4: Memory is updated if the new generated vector of the convergence rate of HS. Hence, a small HMCR value
0
the last step (xj ) is better than the worst vector in HM, based will increase the diversity of the algorithm. PAR enables
on the objective function. modifying the selected value xi,j . Large PAR values will
Step 5: After each improvisation, the algorithm will check improve the intensification ability of HS, whereas small PAR
the stopping criteria, such as the maximum number of impro- values will improve the diversity or exploration of HM. Large
visations, to end the search process. The next pseudocode PAR and HMCR values are better for unimodal problems,
describes the improvisation process of HS algorithm: whereas small values are better for multimodal problems.

VOLUME 7, 2019 14235

 A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

Algorithm 1 Harmony Search Algorithm Improvisation (BW min , BW max ), as shown in Eqs. (4) and (5):
1. while (t < Max number of iterations)  
BW min
 
2. for (j = 1 toD) = {D : number of dimensions} C = ln ÷ NI (4)
3. If (R2) ≤ HMCR {Memory consideration } BW max
4.
0
xi = xi,j {i is a random integer (1, . . . HMS )} BW (t) = BW max × e(c×t) . (5)
5. if (R3 ≤ PAR) {Pitch adjustment } Mahdavi applied the new algorithm on benchmark functions
0 0
6. xj = xj ± R4 × bw and other engineering optimization problems (e.g., minimiza-
7. end if tion of the weight of a spring and pressure vessel design) to
8. else compare the results with the original HS and other optimiza-
0
9. xj = LB + R5 × (UB − LB)) tion algorithms. The new variant provided better results. This
10. end if variant of HS inspired many researchers to develop modified
11. end for versions of the original HS and improve its performance.
12. Update
 0 HM :  The problem with IHS is that it requires selecting the upper

13. if xj better than worst xj xj ∈ HM and lower values of the variable BW within a large range
0
14. xj = xj (0 ∼ ∞).
15. t =t +1
16. End while 2) GLOBAL-BEST HARMONY SEARCH (GHS; 2008)
17. return best harmony Omran et al. [23] proposed a new variant of HS, called GHS.
GHS can handle discrete and continuous cases. It utilizes the
concept of swarm intelligence in particle swarm optimization
(PSO) [27] to replace the BW value. Replacement is accom-
Selecting fixed HMCR and PAR values is difficult because plished by using the best HM value for pitch adjustment
characteristics differ for each optimization problem [25]. instead of BW. The effects of HMS and HMCR were tested on
Different variants of the HS algorithm have been developed the GHS algorithm. A higher value of HMCR was suggested
to cope with various types of parameters, such as binary, dis- to obtain a better result for high-dimensionality problems,
crete, and continuous. Selecting the appropriate values for HS whereas a lower result is better for low-dimensionality prob-
parameters is one of the difficulties in using the HS algorithm lems. The results show that GHS outperforms HS and IHS
and other meta-heuristic algorithms; numerous variants have except at low dimensions. GHS has the same steps as IHS
been created to address this problem by replacing these values except for the pitch-adjusting step. Eq. (6) describes how
with new dynamic values or by adding another component to GHS uses the best result in HM to replace BW values.
solve this issue [26]. In the next sections, the classifications
of some variants and their enhancement of the original HS if (rand ≤ PAR) xj = xbest,n , n ∈ {1, 2, 3, . . . ND} (6)
algorithm are described. Although GHS eliminates the use of the BW value, lower and
upper PAR values should still be specified.
A. CONTINUOUS-BASED HS ALGORITHMS
Different variants of HS have been created to deal with con- 3) A SELF-ADAPTIVE GLOBAL BEST HARMONY
tinuous variable problems. We present some well-known HS SEARCH ALGORITHM (SGHS; 2010)
continuous variants and how they differ from the original HS Pan et al. [25] created a new variant of HS based on GHS [23]
algorithm. for continuous optimization problems. The new variant has a
new improvisation technique and dynamic HMCR and PAR
1) AN IMPROVED HARMONY SEARCH ALGORITHM FOR based on a learning procedure. The BW value is reduced
SOLVING OPTIMIZATION PROBLEMS (IHS; 2007) dynamically with an increasing number of iterations based
Mahdavi et al. [1] developed a dynamic method to gener- on Eq. (7 & 8).
ate PAR and BW values. Eq. (3) describes dynamic PAR as BW max −BW min × 2t NI
BW (t) = BW max − if t < (7)
follows: NI 2
NI
BW (t) = BW min ift ≥ (8) (8)
(PARmax − PARmin ) 2
PAR (t) = PARmin + × t. (3)
NI To update HMCR and PAR values, the initial values of
HMCRm (e.g., 0.98) and PARm (e.g., 0.9) are provided, and
PAR(t) is the value of PAR for each iteration (t), and its HMCRm and PARm values are generated using a normal
value is increased in every iteration based on PARmin and distribution (Gaussian) in every iteration. After a specific
PARmax, which are the minimum and maximum values of number of iterations, such as 100 iterations in this work, the
PAR, respectively. NI denotes the number of iterations. mean values of HMCR and PAR are calculated to obtain a
Meanwhile, BW also has a dynamic value for each iteration successful result, and then used in the next iteration. This
based on the minimum and maximum values of bandwidth step is repeated until the stopping criteria are achieved.

14236 VOLUME 7, 2019

A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

For HMS, the author suggested that the use of small fixed PAR value is decreased within the iterations to achieve
values (5 ∼ 10) is best based on their experiments. great exploitation, as described by Wang and Huang [30]. For
BW, the author borrowed its formula from the IHS [1] variant.
4) NOVEL GLOBAL HARMONY SEARCH ALGORITHM FOR The algorithm was compared with seven previous HS variants
UNCONSTRAINED PROBLEMS (NGHS; 2010) using the benchmark function from CEC05.
Zou et al. [28]. developed another variant that excludes
HMCR and PAR variables and replaces them with position 6) A DIFFERENTIAL-BASED HARMONY SEARCH ALGORITHM
updating and genetic mutation, respectively. The concept of FOR THE OPTIMIZATION of CONTINUOUS PROBLEMS
position updating was borrowed from PSO to prevent the (DH/BEST; 2016)
NGHS algorithm from getting trapped in the local optimum. Abedinpourshotorban et al. [31] introduced a new HS variant
The next pseudocode shows the manner of improvisation of by modifying two aspects of the original HS. The first mod-
NGHS: ification is applied to the initialization of HS by using a new
method to initiate feasible solutions with less randomness.
Algorithm 2 NGHS Improvisation (Zou 2010) The second modification involves replacing pitch adjustment
1. for (j = 1 to D) {D : number of dimensions} with the updated version inspired by the differential evolution
2. XR = 2 × XjBest − Xjworst (DE) mutation strategy and excluding the BW parameter.
3. if (X R > xU ) The following algorithm describes the new initialization
4. XR = xU processes, which is implemented by replacing the random
5. else if (X R < xL ) value with a new calculation based on HMS:
6. XR = xL
7. end if Algorithm 3 DH/Best Initialization (Hosein 2016)
8. x́i = xjworst ± rand (0 ∼ 1) × XR − XjWorst 1. for (j = 1 toD) {D = dimensions
9. if (rand (0 ∼ 1) ≤ Pm) 2. for (i = 1 toHMS)
10. x́i = xL + rand (0 ∼ 1) × (xU − xL ) 3. tempi = LB + ((i − HMS0.5
)) × (UB-LB)
11. end if 4. end for
12. Update The Memory 5. Shuffle the temporary array
13. end for 6. for (i = 1 toHMS)
7. HM = tempi
8. end for
where XjBest and Xjworst are the best and worst Xi in HM,
9. end for
respectively, based on the objective function f (x); xU and xL
are the upper and lower bounds of the objective function,
respectively; and rand (0 ∼ 1) is a random value between where UB and LB are the upper and lower bounds of the deci-
0 and 1. Another modification on the original HS is when the sion variables. The new variant eliminates the requirement of
worst value is updated with the new Xj even if the new value setting BW, and pitches are adjusted based on the distances
is not better than the worst one. between the pitches in HM by using DE/best/1 mutation,
as described in the following pseudocode:
5) AN IMPROVED GLOBAL-BEST HARMONY SEARCH
ALGORITHM (IGHS; 2013) Algorithm 4 DH/Best Improvisation (Hosein 2016)
El-Abd [29] developed an improved variant of GHS by focus- 1. for (i = 1 to D)
ing on the explorative range at the beginning, and then on the 2. if (r(0 ∼ 1) ≤ HMCR)
0
exploitative range at the end of a search. To accomplish this, 3. Xi = Xij (i is random integer from1..HMS)
the author used Gaussian distribution to select the random 4. if (r(0 ∼ 1) ≤ PAR)
0
pitch adjustment, as described in Equation (9): 5. Xi = Xbest + r (0 ∼ 1) × (X r1,J − Xr2,J )
0 0
0
Xj = HM rd + Gauss (0, 1) × BW (9) 6. if (Xj < LBorXj > UB)
0
7. Xj = r (0 ∼ 1) × (UB-LB) + LB
where HM rd is a randomly selected value from HM, and 8. end if
Gauss is a random number with a mean of 0 and a standard 9. end if
deviation of 1. For pitch adjustment, Equation (10) is used as 10. else
0
follows: 11. Xj = r (0 ∼ 1) × (UB-LB) + LB
0 12. end if
Xj = HM best
d + ∅ × BW (10) 13. end for
where HM best
d is the best value in HM based on the objective
function evaluation f (x). The value ∅ is a random number that where UB and LB are the upper and lower bounds of the
is uniformly distributed within the range {−1, 1}. decision variables, r(01) is the random value between 0 and 1,

VOLUME 7, 2019 14237

 A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

Xbest is the best Xi in HM based on the objective function, and optimal adjacent value as follows:
Xr1,J and Xr2,J are two random values in the jth dimension. (
hbj r < PAR
Xij = (13)
B. DISCRETE-BASED HS ALGORITHMS Xi else
Lee et al. [18], Geem [32], and Wang et al. [33] developed
several versions of HS to address the discrete variable prob- where hbj corresponds to the best value within HM (i.e.,
lem. In these previous works, HS was used to solve engineer- global optimal harmony).
ing design problems, which are discrete by nature. In another Two updating techniques: memory consideration strategies
research, Al-Betar used HS to solve another discrete problem, (serial and parallel update mechanisms) and (individual and
which is a university course timetabling optimization prob- bit selection mechanisms) were studied.
lem [14]. Moreover, Gao et al. [34] utilized the HS algorithm Individual selection strategy:
in the no-wait flow shop scheduling problem with total flow (
time criterion. hpj if r1 < HMCR
The improvisation process of discrete HS for Xi Xij = (14)
R else
is described in equation (11), and for continuous in (
equation (12): 0 if r1 < 0.5
R= (15)
0
1 else
xi ← xi (i + c) For discrete design variables (11)
0
xi ← xi + a For continuous design variables (12) where Xij is the jth bit of Xj , the new improvised vector p is the
random value of HMS, and r1 and r2 are two random values
We select the neighbor value by adding c ∈ (−1 ∼ 1) for the between the range of 0 and 1.
discrete problem and a = BW ×u for the continuous problem, The algorithm achieves good results after being tested by
where BW is the bandwidth value and u is a random value benchmark functions and the knapsack problem.
between (−1 ∼ 1).
For the mixed problem between discrete and continuous D. CHAOTIC-BASED HS ALGORITHMS
problems, such as the work of Lee et al. [35], HS is used Chaotic maps are similar to randomness in their function-
in a truss design problem that requires solving both types of ality. However, these algorithms are less sensitive to the
optimization problems (discrete and continuous) in parallel. initial condition and parameters due to their deterministic
HS shows that it can effectively cope with all these types of nature. Chaotic-based HS algorithms are applied to many
problems. meta-heuristic algorithms to improve their performance.
Some HS algorithm variants have been used in chaotic
C. BINARY-BASED HS ALGORITHMS maps [22], [39]–[41]. The following variant describes how
Many researchers have been using HS in discrete and contin- chaotic maps are applied to HS.
uous optimizations in different fields of science and engineer-
ing. Some researchers have utilized HS in binary problems. CHAOTIC HARMONY SEARCH ALGORITHMS (CHS; 2010)
Although HS with binary coding can be used to address Alatas developed a new HS variant by using chaotic maps for
binary-coded problems, the pitch adjustment operator is infe- parameter adoption. the main objective was to avoid falling
rior in binary space, which degrades the performance of the into the local optimum and to improve the convergence of HS.
algorithm because PAR is typically eliminated. For exam- Seven chaotic maps were used to generate random values
ple, Geem [36] utilized HS with binary coding to address instead of the original random generator in HS.
water pump switching problems and the PAR parameter was Maps were used in this work for different parameters of
eliminated in this work. Another utilization of binary HS HS, such as in HM, PAR, and BW generation. Some of the
was performed by Geem and Williams [37] to cope with an maps enhanced solution quality and helped HS escape from
ecological optimization problem that achieved competitive the local optimum. In this work, Alatas used only two bench-
results. The next variant demonstrates how HS binary variants mark functions to demonstrate the capabilities of the new
handle binary problems. variant.

AN IMPROVED BINARY hARMONY SEARCH E. OPPOSITION-BASED HS ALGORITHMS

ALGORITHM (ABHS; 2013) Gao et al. [42] and other researchers used the opposition-
Wang et al. [38] developed a new HS variant to cope with based intelligent computation approach to enhance HS
the binary optimization problem. In the beginning, HM is performance, by creating opposite values from the new
initialized by generating random binary numbers for each improvisation results and replacing the improvisation values
value within HM. Then, a new pitch adjustment is created to if the new results are found to be better. Successful variants
handle the binary problems and improve HS capabilities by that use opposition-based methods are presented in the suc-
replacing the adjacent value in the original HS with a global ceeding subsections.

14238 VOLUME 7, 2019

A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

1) AN IMPROVED GLOBAL BEST HARMONY SEARCH IV. HYBRIDIZATION OF THE HS ALGORITHM

ALGORITHM FOR FASTER OPTIMIZATION (IGHS; 2014) The development of new meta-heuristic algorithms can be
Xiang et al [43] recently developed a new HS variant based accomplished in several ways, and the creation of new
on the GHS [23] algorithm. The new variant exhibits four hybrid algorithms based on existing heuristic algorithms
modifications from the original GHS. First, it uses the opposi- is considered one of the most successful techniques [53].
tion technique (OBL) [44] during initialization to obtain more The best way to create a new successful meta-heuristic
diverse initial values and avoid falling into the local optimum. algorithm is by using an appropriate assembly of existing
Second, the DE concept is adopted to implement new impro- meta-heuristic algorithms, such as combining a multi-agent
visations. Third, HS exploration is improved and falling into algorithm with a trajectory algorithm, to create algorithms
the local minimum is avoided by using two concepts based on with enhanced capabilities [54]. The need to create a multi-
the artificial bee colony algorithm [45] and the OBL scheme. objective meta-heuristic algorithm to handle NP-hard opti-
Fourth, the HMCR and PAR parameters are updated based on mization problems remains a productive research area [4].
the sign and periodic functions. The hybrid algorithms of HS have proven their effective-
Two experimental studies were conducted to compare ness in improving the performance of HS to cope with differ-
IGHS with the original HS and GHS (2008) and other ent cases [55]. The concept of hybridizing the HS algorithm
meta-heuristic algorithms. Competitive results were achieved with other meta-heuristic algorithms has recently become
by IGHS. prevalent, and many hybrid algorithms have been developed
and used [38]. We present well-known hybrid HS categories
2) GLOBAL HARMONY SEARCH WITH GENERALIZED and provide examples for each.
OPPOSITION-BASED LEARNING (GOGHS; 2015)
Gou et al. [46] created a new HS variant because of the poor A. HYBRIDIZATION WITH A LOCAL SEARCH-BASED
exploitation of the original HS. The new variant adopts the ALGORITHM
concept of generalized opposition [47], which is a modified Global search algorithms, such as HS, are well-known for
version of the original opposition [44], in addition to utilizing their good exploration, whereas local search algorithms are
NGHS [28]. appropriate for exploitation. Combining the two types will
GOGHS first initializes its HM in the same manner as the probably produce a strong algorithm for exploration and
original HS. After initialization, a new candidate (Xi ) is cre- exploitation. The following hybrid algorithms are examples
ated using the NGHS algorithm [28]. A modified version of of HS algorithms with a local search algorithm.
(Xi ) is created in the new candidate harmony (O_Xi ) by using
the generalized opposition-based solution. Finally, the worst 1) HYBRIDIZING HARMONY SEARCH ALGORITHM WITH
result in HM will be replaced with the best value between the SEQUENTIAL QUADRATIC PROGRAMMING FOR
new candidate and its opposition (Xi and O_Xi ). ENGINEERING OPTIMIZATION PROBLEMS (HHSA; 2008)
Fesanghary et al. [56] developed an example of a hybrid HS
3) IMPROVED HARMONY SEARCH ALGORITHM: algorithm with a local-based search algorithm. This hybrid
LHS (LHS; 2017) algorithm is incorporated with sequential quadratic program-
Ouyang et al. [48] recently developed an HS variant by ming (SQP) to improve the local search speed and accuracy of
implementing three modifications to the original HS to the HS algorithm. The hybrid algorithm achieves better result
improve its effectiveness. First, they used the opposition compared with the original HS and other heuristic algorithms.
concept [44], within the harmony consideration process The effects of HS parameters, such as HMCR, HMS, and
to obtain better diversity in the result. Second, they used PAR, were also tested in this work. A small HMCR value
self-adaptation in pitch adjustment to improve the inten- will reduce the algorithm’s efficiency, whereas a large value
sification of the HS algorithm by mimicking the swarm will result in better exploration. However, this condition may
intelligence concept of PSO. Third, they found the worst cause the algorithm to fall into the local optimum. For HMS,
vector in HM and updated it with the best value between the larger the size, the better the results obtained with numer-
the improvisation and opposition results. Finally, they com- ous iterations. However, this process can be time-consuming.
pared the new variant with 10 previous HS variants using By contrast, a small HMS value can cause premature con-
48 mathematical benchmark functions. The new variant vergence. The authors suggest using an HMS value between
exhibited higher competitive performance than the previous 4 and 10. For PAR, a large value with a small BW will
variants. increase HS efficiency.

F. MULTI-OBJECTIVE HS ALGORITHM 2) HHS, HGHS, HMGHS (2010)

After the success of applying HS to single-objective opti- Wang et al. [33] made another hybrid local/global search
mization problems, many researchers have applied HS to algorithm in 2010, and developed three hybrid HS algo-
multi-objective optimization problems [49]–[52]. rithms to reduce the total time of flow shop scheduling

VOLUME 7, 2019 14239

 A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

in a blocking optimization problem. The first algorithm is V. APPLICATIONS OF HS ALGORITHM

based on the original HS by Geem et al. [7], the second is Multiple mathematical techniques, such as dynamic, lin-
based on GHS [23], and the third is based on a modified ear, and nonlinear programming, have been used to solve
GHS. The authors hybridized the HS variants with a local optimization problems before the existence of heuristic algo-
search algorithm based on the insert neighborhood to bal- rithms. All traditional techniques can find the global opti-
ance the exploration and exploitation of the new algorithm. mal solution, but special specifications or characteristics
They implemented HM initialization using another heuristic should be specified for each problem, and thus, covering
algorithm (NEH) [57]. Furthermore, they transformed the NP-hard problems is impossible. Another disadvantage of
continuous vectors of harmonies to discrete job adjustments these methods is that they consume considerable space and
based on the largest position value. To test the efficiency time in solving a problem, thereby increasing the complex-
and effectiveness of their hybrid algorithms, the authors used ity of the problem and making these methods impracti-
the blocking flow shop scheduling problem, which is consid- cal for real-world problems. Consequently, meta-heuristic
ered an NP-hard problem when more than two machines are algorithms have been developed by mimicking nature or
involved. human intelligence to solve these issues. These algorithms
have proven their efficiency in handling real NP-hard
B. HYBRIDIZATION WITH POPULATION-BASED problems [66].
ALGORITHM HS was initially used to find the best design for water
Another way to improve population search algorithms is by distribution networks by its original author Geem There-
creating a hybrid algorithm with another population search after, many researchers have utilized HS in different fields,
algorithm (or some of its components) to combine the advan- such as engineering and information technology. Several
tages of both algorithms. The next hybrid algorithm is an categories of these works are described in the subsequent
example of a hybrid population-based algorithm. sections.

AN EFFECTIVE HYBRID HARMONY SEARCH-BASED A. REAL-WORLD APPLICATIONS /INDUSTRIAL

ALGORITHM FOR SOLVING MULTIDIMENSIONAL APPLICATIONS
KNAPSACK PROBLEMS (2015) Many researchers have used HS to solve real-world issues.
Zhang et al. [58] developed a new hybrid algorithm of HS Music composition, an example of a real-world applica-
with fruit fly optimization (FFO) [59]. The main objective of tion, has been conducted to support composers, such as
this algorithm is to overcome the weak exploitation process Navarro et al. [67].
of HS. This hybrid algorithm is used to solve MKPs. Al-Betar et al. [2], [15] and Chiarandini et al. [13] used
Binary string values are used in this work to apply the HS and its hybrids to create a university course timetable,
new hybrid algorithm to MKPs, with few modifications to the which is considered a real-world optimization problem. Many
original HS algorithm. The hybrid algorithm starts by using researchers have applied HS to solve real-world problems,
the original HS initialization process. Then, the global HS such as tour planning, logistics, and project and flow shop
scheme is used to improvise the values in HM that are similar scheduling, as shown in Table 1.
to GHS [23]. The smell-based FFO search is used to create
modified values of HM and update each value of HM to see if
each value is better than the original. A repair operator is used B. COMPUTER SCIENCE PROBLEMS
to ensure that the new improvised values satisfy the weight Different applications in computer science have been solved
constraints. using the HS algorithm, such as web page and document
clustering. Clustering refers to the automatic classification
C. HYBRIDIZATION HS WITH OTHER COMPONENTS of a webpage or document content to reduce the amount of
The use of HS or its component to assist another meta- data and gain useful information. Forsati et al. [77], [78]
heuristic algorithm through hybridization has been realized in used a hybrid HS with K-means to cluster web pages.
different studies [60]–[63]. The subsequent hybrid algorithm Cobos et al. [79] used GHS [23] with K-means, frequent
is an example of using HS as a component in the search term sets, and the Bayesian information criterion to cluster
process. webpage data. Text summarization is another computer appli-
cation of text extraction that aims to summarize the text of
PSO, ANT COLONY OPTIMIZATION (ACO), AND HS SCHEME web content, such that the end user can easily find what he/she
HYBRIDIZED TO OPTIMIZE TRUSS STRUCTURES is looking for in summarized form. Different studies have
Kaveh et al. [64] proposed a new hybrid HS algorithm with been conducted using HS and its variant GHS to accomplish
PSO [27] and ACO [65]. HHS, HGHS, and HMGHS used text summarization [47], [80], [81].
the hybrid principle to solve the continuous and truss design Other computer applications of HS are used in the liter-
optimization problems. In this work, the role of HS was to ature in other areas, such as Internet routing, visual track-
control the variable constraints through the design of the truss ing, robotics, and automatic software test case generation,
structure optimization problem. as shown in Table 1.

14240 VOLUME 7, 2019

A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

TABLE 1. Applications of HS. production while meeting power requirements. A new con-
cept, namely, combined heat and power (CHP), was formed
to achieve maximum efficiency because the primitive pro-
duction of electricity with the use of fossil fuels can only
achieve up to 60% efficiency, thereby causing wasted heat.
However, balancing heat and power produced by the same
units is required. Vasebi applied HS to solve the CHP eco-
nomic dispatch problem by using two cases and compared
the results with those of other meta-heuristic algorithms.
Since then, other researchers have used HS to solve CHP
problems [94], [95].
Table 1 shows that considerable research has utilized
HS to optimize power streaming systems by achieving the
minimum cost to transfer electricity in a power system.
HS optimizes voltage nodes as a variable for a set of non-
linear equations as an objective function. Sivasubramani and
Swarup [49] developed multi-objective HS to minimize fuel
consumption, which will reduce the cost and power transfer
emission, which will then reduce energy loss. The results
obtained were compared with those of GA. The HS results
outperform those of GA.
HS is utilized in different fields of electric engineering
research, such as optimization for renewable energy charg-
ing [101], wind power generation [133], and multilevel
inverter optimization [100].

D. CIVIL ENGINEERING PROBLEMS

Considerable research has been conducted on structural
design optimization because it is considered a sensitive and
difficult activity, particularly when time and cost are consid-
ered. Lee and Geem [10] used HS to cope with this optimiza-
tion problem, and the result outperformed those of traditional
techniques, such as linear and nonlinear equations, because
unlike traditional techniques, HS does not require a high
volume of gradient calculation. References [104] and [134]
also used HS to cope with this problem.

E. MECHANICAL ENGINEERING PROBLEMS

Several works have utilized HS to cope with mechanical engi-
neering problems. Zarei et al. [121] used the HS algorithm
to find the ideal cutting parameters of a multi-pass face-
milling machine, thereby improving production quality and
minimizing the cost and time consumption. Different param-
eters have been optimized using HS to improve machine
tool capabilities, such as the number of passes, the depth of
cut for each pass, speed, and feed. The result showed better
performance of HS compared with that of GA.
Another mechanical problem solved by HS is the schedul-
ing of diesel generators in oil rig platforms, as reported by
Yadav et al. [119]. The author attempted to reduce the fuel
consumption of a diesel generator that supports the oil rig,
C. ELECTRICAL ENGINEERING PROBLEMS while meeting load demand and operational constraints to
Research on power dispatch and energy flow manage- achieve the optimum performance of the rig. The improved
ment systems has been utilizing HS in different works. variant of HS obtained competitive results and convergence
Vasebi et al. [93] used HS to find the minimum cost of energy with the original HS and other meta-heuristic algorithms.

VOLUME 7, 2019 14241

 A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

F. BIOLOGICAL AND MEDICAL APPLICATIONS • HS gets stuck in a local search when applied to numeri-
Forecasting infectious diseases is an interesting idea, and we cal applications.
can formulate strategies to minimize them by predicting their
effects. A study was conducted by Hickmann et al. [126] on REFERENCES
infectious disease forecasting using HS. Data were obtained [1] M. Mahdavi, M. Fesanghary, and E. Damangir, ‘‘An improved harmony
from the Centers for Disease Control and Prevention and search algorithm for solving optimization problems,’’ Appl. Math. Com-
put., vol. 188, no. 2, pp. 1567–1579, May 2007.
Wikipedia article access logs, and then analyzed using epi- [2] M. A. Al-Betar and A. T. Khader, ‘‘A harmony search algorithm for
demiological models to predict the effect of infectious dis- University course timetabling,’’ Ann. Oper. Res., vol. 194, no. 3, pp. 3–31,
eases in the United States. The HS algorithm was used in the 2012.
prediction process to solve a nonlinear equation. [3] K. S. Lee and Z. W. Geem, ‘‘A new meta-heuristic algorithm for con-
tinuous engineering optimization: Harmony search theory and practice,’’
The HS algorithm was also used to optimize radioisotope Comput. Methods Appl. Mech. Eng., vol. 194, no. 3, pp. 3902–3933, 2005.
placement and intensities by Panchal [131]. Radiation is used [4] X.-S. Yang, ‘‘Harmony search as a metaheuristic algorithm,’’ in Music-
to diagnose and treat cancer as part of medical physics. Inspired Harmony Search Algorithm. Springer, 2009, pp. 1–14.
[5] K. Sörensen, ‘‘Metaheuristics—The metaphor exposed,’’ Int. Trans. Oper.
GA and SA were applied to optimize the radioisotope, but Res., vol. 22, no. 3, pp. 3–18, 2015.
the results were obtained faster by using HS compared with [6] S. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, ‘‘Optimization by simu-
GA, thereby improving the patient treatment process. lated annealing,’’ Science, vol. 220, no. 4598, pp. 671–680, 1983.
[7] Z. W. Geem, J. H. Kim, and G. V. Loganathan, ‘‘A new heuristic opti-
Table 1 shows the full categories of HS applications with
mization algorithm: Harmony search,’’ J. Simul., vol. 76, no. 2, pp. 60–68,
related research works. Feb. 2001.
[8] R. Storn and K. Price, ‘‘Differential evolution-a simple and efficient
heuristic for global optimization over continuous spaces,’’ J. Global
VI. DISCUSSION AND CONCLUSION Optim., vol. 11, no. 4, pp. 341–359, 1997.
This article describes the HS optimization algorithm, some of [9] F. Glover, ‘‘Heuristics for integer programming using surrogate con-
its variants, applications, and hybrid versions. The algorithm straints,’’ Decis. Sci., vol. 8, no. 1, pp. 156–166, 1977.
[10] K. S. Lee and Z. W. Geem, ‘‘A new structural optimization method based
has maintained considerable interest since its inception until on the harmony search algorithm,’’ Comput. Struct., vol. 82, nos. 9–10,
now because of its ability to balance between exploration and pp. 781–798, 2004.
exploitation, and its ability to be used in different research [11] A. E. Eiben and C. A. Schippers, ‘‘On evolutionary exploration and
exploitation,’’ Fundam. Inform., vol. 35, nos. 1–4, pp. 35–50, 1998.
areas because of its simplicity. The author of HS exerted con- [12] K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, ‘‘A fast and elitist
siderable effort to utilize the algorithm in different research multiobjective genetic algorithm: NSGA-II,’’ IEEE Trans. Evol. Comput.,
areas, and a webpage that contains the full information of HS vol. 6, no. 2, pp. 182–197, Apr. 2002.
[13] M. Chiarandini, M. Birattari, K. Socha, and O. Rossi-Doria, ‘‘An effective
is available for the benefit of other researchers. hybrid algorithm for university course timetabling,’’ J. Scheduling, vol. 9,
We noticed considerable utilization of HS and improve- no. 5, pp. 403–432, 2006.
ment on HS. However, the utilization and improvement of HS [14] M. A. Al-Betar, A. T. Khader, and I. Y. Liao, ‘‘A harmony search
continues, particularly in terms of its hybridization with other with multi-pitch adjusting rate for the University course timetabling,’’
in Recent Advances in Harmony Search Algorithm. Springer, 2010,
meta-heuristic algorithms. A standard framework or criterion pp. 147–161.
to compare meta-heuristic algorithms or its variants in the [15] M. A. Al-Betar and A. T. M. Khader Zaman, ‘‘University course
meta-heuristic field is necessary. timetabling using a hybrid harmony search metaheuristic algorithm,’’
IEEE Trans. Syst., Man, Cybern. C, Appl. Rev., vol. 42, no. 5,
Finally, the following are the advantages of HS over previ- pp. 664–681, Sep. 2012.
ous optimization algorithms: [16] S. O. Degertekin, ‘‘Optimum design of steel frames using harmony search
algorithm,’’ Struct. Multidiscipl. Optim., vol. 36, no. 3, pp. 393–401,
• HS has proven its ability to be applied to different fields 2008.
of real applications [66]. [17] Z. W. Geem, Harmony Search Algorithms for Structural Design Opti-
mization, vol. 239. Berlin, Germany: Springer, 2009.
• HS can cover discrete or continuous optimization prob- [18] K. S. Lee, Z. W. Geem, S.-H. Lee, and K.-W. Bae, ‘‘The harmony search
lems, and decimal and binary alphabets [66]. heuristic algorithm for discrete structural optimization,’’ Eng. Optim.,
• HS can distinguish the high-performance zone in the vol. 37, no. 7, pp. 663–684, 2005.
solution range within a short time. [19] Z. W. Geem, ‘‘Optimal cost design of water distribution networks using
harmony search,’’ Eng. Optim., vol. 38, no. 3, pp. 259–277, 2006.
• HS requires less mathematical operations. [20] Z. W. Geem, ‘‘Particle-swarm harmony search for water network design,’’
• HS does not exhibit divergence. Eng. Optim., vol. 41, no. 3, pp. 297–311, 2009.
• Constructive information of the optimization problem is [21] Z. W. Geem, J. H. Kim, and G. Loganathan, ‘‘Harmony search optimiza-
tion: Application to pipe network design,’’ Int. J. Model. Simul., vol. 22,
not important in HS because it uses random search. no. 3, pp. 125–133, 2002.
• HS produces a new solution in accordance with all pre- [22] B. Alatas, ‘‘Chaotic harmony search algorithms,’’ Appl. Math. Comput.,
vious solutions and simultaneously covers all vectors. vol. 216, no. 3, pp. 2687–2699, 2010.
[23] M. G. H. Omran and M. Mahdavi, ‘‘Global-best harmony search,’’ Appl.
• HS is easy to program, and its simplicity makes it easy
Math. Comput., vol. 198, no. 2, pp. 643–656, 2008.
to combine with other meta-heuristic algorithms. [24] M. A. Al-Betar, I. A. Doush, A. T. Khader, and M. A. Awadallah, ‘‘Novel
selection schemes for harmony search,’’ Appl. Math. Comput., vol. 218,
The weaknesses of the HS algorithm are as follows: no. 10, pp. 6095–6117, 2012.
[25] Q.-K. Pan, P. N. Suganthan, M. F. Tasgetiren, and J. J. Liang, ‘‘A self-
• HS is highly sensitive to its parameter values, such as adaptive global best harmony search algorithm for continuous optimiza-
HMCR, PAR, and BW [30]. tion problems,’’ Appl. Math. Comput., vol. 216, no. 3, pp. 830–848, 2010.

14242 VOLUME 7, 2019

A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

[26] O. Moh’d Alia and R. Mandava, ‘‘The variants of the harmony search [50] S. Sivasubramani and K. Swarup, ‘‘Environmental/economic dispatch
algorithm: An overview,’’ Artif. Intell. Rev., vol. 36, no. 1, pp. 49–68, using multi-objective harmony search algorithm,’’ Electr. Power Syst.
2011. Res., vol. 81, no. 9, pp. 1778–1785, 2011.
[27] R. Eberhart and J. Kennedy, ‘‘A new optimizer using particle swarm [51] K. Nekooei, M. M. Farsangi, H. Nezamabadi-Pour, and K. Y. Lee,
theory,’’ in Proc. 6th Int. Symp. Micro Mach. Hum. Sci. (MHS), Oct. 1995, ‘‘An improved multi-objective harmony search for optimal placement of
pp. 39–43. DGs in distribution systems,’’ IEEE Trans. Smart Grid, vol. 4, no. 1,
[28] D. Zou, L. Gao, J. Wu, and S. Li, ‘‘Novel global harmony search pp. 557–567, Mar. 2013.
algorithm for unconstrained problems,’’ Neurocomputing, vol. 73, no. 3, [52] K.-Z. Gao, P. N. Suganthan, Q.-K. Pan, T. J. Chua, T. X. Cai, and
pp. 3308–3318, 2010. C.-S. Chong, ‘‘Pareto-based grouping discrete harmony search algorithm
[29] M. El-Abd, ‘‘An improved global-best harmony search algorithm,’’ Appl. for multi-objective flexible job shop scheduling,’’ Inf. Sci., vol. 289,
Math. Comput., vol. 222, pp. 94–106, Oct. 2013. pp. 76–90, Dec. 2014.
[30] C.-M. Wang and Y.-F. Huang, ‘‘Self-adaptive harmony search algorithm [53] E. Corchado and A. A. A. de Carvalho, ‘‘Hybrid intelligent algorithms
for optimization,’’ Expert Syst. Appl., vol. 37, no. 3, pp. 2826–2837, 2010. and applications,’’ Inf. Sci., Int. J., vol. 180, no. 14, pp. 2633–2634, 2010.
[31] H. Abedinpourshotorban, S. Hasan, S. M. Shamsuddin, and [54] A. Alaa, A. R. A. M. Al-Sewari, B. Ameen, S. Hammoudeh, and
N. F. As’Sahra, ‘‘A differential-based harmony search algorithm for Z. Kamal, ‘‘Hyperdize jaya algorithm for harmony search algo-
the optimization of continuous problems,’’ Expert Syst. Appl., vol. 62, rithm’s parameters selection,’’ Univ. Malaysia Pahang, Pekan, Malaysia,
pp. 317–332, Nov. 2016. Tech. Rep., 2016, pp. 788–791.
[32] Z. W. Geem, ‘‘Novel derivative of harmony search algorithm for discrete [55] A. Alomoush, A. A. Alsewari, H. S. Alamri, and K. Z. Zamli, ‘‘Solving
design variables,’’ Appl. Math. Comput., vol. 199, no. 1, pp. 223–230, 0/1 knapsack problem using hybrid HS and Jaya algorithms,’’ Adv. Sci.
2008. Lett., vol. 24, no. 3, pp. 7486–7489, 2018.
[33] L. Wang, Q.-K. Pan, and M. F. Tasgetiren, ‘‘Minimizing the total flow [56] M. Fesanghary, M. Mahdavi, M. Minary-Jolandan, and Y. Alizadeh,
time in a flow shop with blocking by using hybrid harmony search ‘‘Hybridizing harmony search algorithm with sequential quadratic pro-
algorithms,’’ Expert Syst. Appl., vol. 37, no. 12, pp. 7929–7936, 2010. gramming for engineering optimization problems,’’ Comput. Methods
[34] K.-Z. Gao, Q.-K. Pan, and J.-Q. Li, ‘‘Discrete harmony search algorithm Appl. Mech. Eng., vol. 197, no. 3, pp. 3080–3091, 2008.
for the no-wait flow shop scheduling problem with total flow time crite- [57] M. Nawaz, E. E. Enscore, Jr., and I. Ham, ‘‘A heuristic algorithm for the
rion,’’ Int. J. Adv. Manuf. Technol., vol. 56, nos. 5–8, pp. 683–692, 2011. m-machine, n-job flow-shop sequencing problem,’’ Omega, vol. 11, no. 3,
[35] K. S. Lee, S. W. Han, and Z. W. Geem, ‘‘Discrete size and discrete- pp. 91–95, 1983.
continuous configuration optimization methods for truss structures using [58] B. Zhang, Q.-K. Pan, X.-L. Zhang, and P.-Y. Duan, ‘‘An effective hybrid
the harmony search algorithm,’’ Int. J. Optim. Civil Eng., vol. 1, harmony search-based algorithm for solving multidimensional knapsack
pp. 107–126, Jan. 2011. problems,’’ Appl. Soft Comput., vol. 29, pp. 288–297, Apr. 2015.
[59] W. T. Pan, ‘‘A new evolutionary computation approach: Fruit fly opti-
[36] Z. W. Geem, ‘‘Harmony search in water pump switching problem,’’ in
mization algorithm,’’ in Proc. Conf. Digit. Technol. Innov. Manage., 2011,
Proc. Int. Conf. Natural Comput., 2005, pp. 751–760.
pp. 382–391.
[37] Z. W. Geem and J. C. Williams, ‘‘Ecological optimization using har-
[60] H. Moeinzadeh, E. Asgarian, M. Zanjani, A. Rezaee, and M. Seidi,
mony search,’’ in Proc. Amer. Conf. Appl. Math., World Sci. Eng. Acad.
‘‘Combination of harmony search and linear discriminate analysis
Soc. (WSEAS), 2008, pp. 148–152.
to improve classification,’’ in Proc. 3rd Asia Int. Conf. Modelling
[38] L. Wang, R. Yang, Y. Xu, Q. Niu, P. M. Pardalos, and M. Fei,
Simulation (AMS), 2009, pp. 131–135.
‘‘An improved adaptive binary harmony search algorithm,’’ Inf. Sci.,
[61] F. Nadi, A. T. Khader, and M. A. Al-Betar, ‘‘Adaptive genetic algorithm
vol. 232, pp. 58–87, May 2013.
using harmony search,’’ in Proc. 12th Annu. Conf. Genet. Evol. Comput.,
[39] O. Abdel-Raouf, I. El-Henawy, and M. Abdel-Baset, ‘‘A novel hybrid 2010, pp. 819–820.
flower pollination algorithm with chaotic harmony search for solving
[62] M. J. Li, M. K. Ng, Y.-M. Cheung, and J. Z. Huang, ‘‘Agglomerative fuzzy
sudoku puzzles,’’ Int. J. Mod. Educ. Comput. Sci., vol. 6, no. 3, p. 38,
K-means clustering algorithm with selection of number of clusters,’’ IEEE
2014.
Trans. Knowl. Data Eng., vol. 20, no. 11, pp. 1519–1534, Nov. 2008.
[40] M. F. El-Santawy, A. N. Ahmed, Z. El-Dean, and A. Ramadan, ‘‘Chaotic [63] Q. Li, N. Mitianoudis, and T. Stathaki, ‘‘Spatial kernel K-harmonic means
Harmony Search Optimizer for Solving Numerical Integration,’’ Comput. clustering for multi-spectral image segmentation,’’ IET Image Process.,
Inf. Syst., vol. 16, no. 2, pp. 21–24, 2012. vol. 1, no. 3, pp. 156–167, 2007.
[41] O. Abdel-Raouf, I. El-henawy, and M. Abdel-Baset, ‘‘Chaotic harmony [64] A. Kaveh and S. Talatahari, ‘‘Particle swarm optimizer, ant colony strat-
search algorithm with different chaotic maps for solving assignment egy and harmony search scheme hybridized for optimization of truss
problems,’’ Int. J. Comput. Appl., vol. 86, no. 10, pp. 1–6, 2014. structures,’’ Comput. Struct., vol. 87, no. 3, pp. 267–283, 2009.
[42] X. Z. Gao, X. Wang, S. J. Ovaska, and K. Zenger, ‘‘A hybrid optimization [65] M. Dorigo, M. Birattari, and T. Stutzle, ‘‘Ant colony optimization,’’ IEEE
method of harmony search and opposition-based learning,’’ Eng. Optim., Comput. Intell. Mag., vol. 1, no. 4, pp. 28–39, Nov. 2006.
vol. 44, no. 8, pp. 895–914, 2012. [66] D. Manjarres et al., ‘‘A survey on applications of the harmony search
[43] W.-L. Xiang, M.-Q. An, Y.-Z. Li, R.-C. He, and J.-F. Zhang, algorithm,’’ Eng. Appl. Artif. Intell., vol. 26, no. 8, pp. 1818–1831,
‘‘An improved global-best harmony search algorithm for faster optimiza- 2013.
tion,’’ Expert Syst. Appl., vol. 41, no. 13, pp. 5788–5803, Oct. 2014. [67] M. Navarro, J. M. Corchado, and Y. Demazeau, ‘‘MUSIC-MAS: Mod-
[44] H. R. Tizhoosh, ‘‘Opposition-based learning: A new scheme for machine eling a harmonic composition system with virtual organizations to
intelligence,’’ in Proc. Int. Conf. Comput. Intell. Model., Control assist novice composers,’’ Expert Syst. Appl., vol. 57, pp. 345–355,
Automat., Int. Conf. Intell. Agents, Web Technol. Internet Commerce, Sep. 2016.
Nov. 2005, pp. 695–701. [68] Z. W. Geem, C.-L. Tseng, and Y. Park, ‘‘Harmony search for generalized
[45] D. Karaboga and B. Basturk, ‘‘Artificial bee colony (ABC) optimization orienteering problem: Best touring in China,’’ in Proc. Int. Conf. Natural
algorithm for solving constrained optimization problems,’’ in Proc. Int. Comput., 2005, pp. 741–750.
Fuzzy Syst. Assoc. World Congr., 2007, pp. 789–798. [69] G. Bo, M. Huang, W. H. Ip, and X. Wang, ‘‘The harmony search for
[46] Z. Guo, S. Wang, X. Yue, and H. Yang, ‘‘Global harmony search with the routing optimization in fourth party logistics with time windows,’’
generalized opposition-based learning,’’ Soft Comput., vol. 21, no. 3, in Proc. IEEE Congr. Evol. Comput. (CEC), May 2009, pp. 962–967.
pp. 2129–2137, 2017. [70] M. Huang, G. Bo, X. Wang, and W. H. Ip, ‘‘The optimization of routing in
[47] H. Wang, Z. Wu, Y. Liu, J. Wang, D. Jiang, and L. Chen, ‘‘Space fourth-party logistics with soft time windows using harmony search,’’ in
transformation search: A new evolutionary technique,’’ in Proc. 1st Proc. 6th Int. Conf. Natural Comput. (ICNC), Aug. 2010, pp. 4344–4348.
ACM/SIGEVO Summit Genet. Evol. Comput., 2009, pp. 537–544. [71] Z. W. Geem, ‘‘Multiobjective optimization of time-cost trade-off
[48] H. Ouyang, L. Gao, S. Li, X. Kong, Q. Wang, and D. X. Zou, using harmony search,’’ J. Construct. Eng. Manage., vol. 136, no. 3,
‘‘Improved harmony search algorithm: LHS,’’ Appl. Soft Comput., vol. 53, pp. 711–716, 2009.
pp. 133–167, Apr. 2017. [72] F. De Paola, M. Giugni, F. Pugliese, and P. Romano, ‘‘Optimal design
[49] S. Sivasubramani and K. Swarup, ‘‘Multi-objective harmony search algo- of LIDs in urban stormwater systems using a harmony-search decision
rithm for optimal power flow problem,’’ Int. J. Elect. Power Energy Syst., support system,’’ Water Resour. Manage., vol. 32, no. 3, pp. 4933–4951,
vol. 33, no. 3, pp. 745–752, 2011. 2018.

VOLUME 7, 2019 14243

 A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

[73] Y. Yuan, H. Xu, and J. Yang, ‘‘A hybrid harmony search algorithm for the [96] E. E. Elattar, ‘‘Modified harmony search algorithm for combined eco-
flexible job shop scheduling problem,’’ Appl. Soft Comput., vol. 13, no. 3, nomic emission dispatch of microgrid incorporating renewable sources,’’
pp. 3259–3272, 2013. Energy, vol. 159, pp. 496–507, Sep. 2018.
[74] K.-Z. Gao, P. N. Suganthan, Q. K. Pan, T. J. Chua, T. X. Cai, and [97] A. U. Rehman et al., ‘‘Efficient energy management system using firefly
C. S. Chong, ‘‘Discrete harmony search algorithm for flexible job shop and harmony search algorithm,’’ in Proc. Int. Conf. Broadband Wireless
scheduling problem with multiple objectives,’’ J. Intell. Manuf., vol. 27, Comput., Commun. Appl., 2017, pp. 37–49.
no. 2, pp. 363–374, 2016. [98] Z. W. Geem and H. Hwangbo, ‘‘Application of harmony search to multi-
[75] Y. Yuan and H. Xu, ‘‘An integrated search heuristic for large-scale flexible objective optimization for satellite heat pipe design,’’ in Proc. UKC, 2006,
job shop scheduling problems,’’ Comput. Oper. Res., vol. 40, no. 12, pp. 1–3.
pp. 2864–2877, 2013. [99] M. Fesanghary, E. Damangir, and I. Soleimani, ‘‘Design optimization
[76] Y. Y. Moon, Z. W. Geem, and G.-T. Han, ‘‘Vanishing point detection for of shell and tube heat exchangers using global sensitivity analysis
self-driving car using harmony search algorithm,’’ Swarm Evol. Comput., and harmony search algorithm,’’ Appl. Therm. Eng., vol. 29, nos. 5–6,
vol. 41, pp. 111–119, Aug. 2018. pp. 1026–1031, 2009.
[77] R. Forsati, M. Mahdavi, M. Kangavari, and B. Safarkhani, ‘‘Web page [100] B. Majidi, H. R. Baghaee, G. B. Gharehpetian, J. Milimonfared, and
clustering using harmony search optimization,’’ in Proc. Can. Conf. Elect. M. Mirsalim, ‘‘Harmonic optimization in multi-level inverters using har-
Comput. Eng. (CCECE), May 2008, pp. 1601–1604. mony search algorithm,’’ in Proc. IEEE 2nd Int. Power Energy Conf.
[78] R. Forsati, M. Meybodi, M. Mahdavi, and A. Neiat, ‘‘Hybridization of (PECon), Dec. 2008, pp. 646–650.
K-means and harmony search methods for Web page clustering,’’ in [101] Z. W. Geem and Y. Yoon, ‘‘Harmony search optimization of renewable
Proc. IEEE/WIC/ACM Int. Conf. Web Intell. Intell. Agent Technol., vol. 1, energy charging with energy storage system,’’ Int. J. Elect. Power Energy
Dec. 2008, pp. 329–335. Syst., vol. 86, pp. 120–126, Mar. 2017.
[79] C. Cobos, J. Andrade, W. Constain, M. Mendoza, and E. León, ‘‘Web [102] H. Dong, Y. Bo, M. Gao, and T. Zhu, ‘‘Improved harmony search for
document clustering based on global-best harmony search, K-means, detection with photon density wave,’’ in Proc. Int. Symp. Photoelectron.
frequent term sets and Bayesian information criterion,’’ in Proc. IEEE Detection Imag., Technol. Appl., 2007, p. 662523.
Congr. Evol. Comput. (CEC), Jul. 2010, pp. 1–8. [103] R. Zhang and L. Hanzo, ‘‘Iterative multiuser detection and channel decod-
[80] E. Shareghi and L. S. Hassanabadi, ‘‘Text summarization with harmony ing for DS-CDMA using harmony search,’’ IEEE Signal Process. Lett.,
search algorithm-based sentence extraction,’’ in Proc. 5th Int. Conf. Soft vol. 16, no. 10, pp. 917–920, Oct. 2009.
Comput. Transdiscipl. Sci. Technol., 2008, pp. 226–231. [104] M. P. Saka, ‘‘Optimum design of steel sway frames to BS5950 using
[81] M. Mendoza, C. Cobos, and E. León, ‘‘Extractive single-document sum- harmony search algorithm,’’ J. Construct. Steel Res., vol. 65, no. 3,
marization based on global-best harmony search and a greedy local pp. 36–43, 2009.
optimizer,’’ in Proc. Mex. Int. Conf. Artif. Intell., 2015, pp. 52–66. [105] T. García-Segura, V. Yepes, and J. A. González, ‘‘Computer-support tool
[82] R. Forsati, A. Haghighat, and M. Mahdavi, ‘‘Harmony search based algo- to optimize bridges automatically,’’ Int. J. Comput. Methods Exp. Meas.,
rithms for bandwidth-delay-constrained least-cost multicast routing,’’ vol. 5, no. 3, pp. 171–178, 2017.
Comput. Commun., vol. 31, no. 10, pp. 2505–2519, 2008. [106] Z. W. Geem, ‘‘Hydraulic analysis of water distribution network using
harmony search,’’ in Proc. World Environ. Water Resour. Congr., 2009,
[83] H. Karahan, G. Gurarslan, and Z. W. Geem, ‘‘Parameter estimation of
pp. 1–9.
the nonlinear Muskingum flood-routing model using a hybrid harmony
search algorithm,’’ J. Hydrol. Eng., vol. 18, no. 3, pp. 352–360, 2012. [107] Z. W. Geem, ‘‘Optimal scheduling of multiple dam system using harmony
search algorithm,’’ in Proc. Int. Work-Conf. Artif. Neural Netw., 2007,
[84] B. Zeng and Y. Dong, ‘‘An improved harmony search based energy-
pp. 316–323.
efficient routing algorithm for wireless sensor networks,’’ Appl. Soft
Comput., vol. 41, pp. 135–147, Apr. 2016. [108] Z. W. Geem, ‘‘Parameter estimation of the nonlinear Muskingum model
using parameter-setting-free harmony search,’’ J. Hydrol. Eng., vol. 16,
[85] J. Fourie, S. Mills, and R. Green, ‘‘Harmony filter: A robust visual
no. 3, pp. 684–688, 2010.
tracking system using the improved harmony search algorithm,’’ Image
[109] J. H. Kim, Z. W. Geem, and E. S. Kim, ‘‘Parameter estimation of the
Vis. Comput., vol. 28, no. 3, pp. 1702–1716, 2010.
nonlinear Muskingum model using harmony search,’’ JAWRA J. Amer.
[86] E. Yazdi and A. T. Haghighat. (2010). ‘‘Evolution of biped walking using
Water Resour. Assoc., vol. 37, no. 3, pp. 1131–1138, 2001.
neural oscillators controller and harmony search algorithm optimizer.’’
[110] M. T. Ayvaz, ‘‘Application of harmony search algorithm to the solution
[Online]. Available: https://arxiv.org/abs/1006.4553
of groundwater management models,’’ Adv. Water Resour., vol. 32, no. 3,
[87] A. R. A. Alsewari and K. Z. Zamli, ‘‘A harmony search based pairwise pp. 916–924, 2009.
sampling strategy for combinatorial testing,’’ Int. J. Phys. Sci., vol. 7,
[111] Y. M. Cheng, L. Li, T. Lansivaara, S. C. Chi, and Y. J. Sun, ‘‘An improved
no. 7, pp. 1062–1072, 2012.
harmony search minimization algorithm using different slip surface gen-
[88] A. R. A. Alsewari and K. Z. Zamli, ‘‘Design and implementation of eration methods for slope stability analysis,’’ Eng. Optim., vol. 40, no. 2,
a harmony-search-based variable-strength t-way testing strategy with pp. 95–115, 2008.
constraints support,’’ Inf. Softw. Technol., vol. 54, no. 6, pp. 553–568,
[112] E. T. Yassen, M. Ayob, M. Z. A. Nazri, and N. R. Sabar, ‘‘Meta-harmony
2012.
search algorithm for the vehicle routing problem with time windows,’’
[89] A. A. Alsewari and K. Z. Zamli, ‘‘Interaction test data generation using Inf. Sci., vol. 325, pp. 140–158, Dec. 2015.
harmony search algorithm,’’ in Proc. IEEE Symp. Ind. Electron. Appl. [113] R. Gholizadeh, G. G. Amiri, and B. Mohebi, ‘‘An alternative approach to
(ISIEA), Sep. 2011, pp. 559–564. a harmony search algorithm for a construction site layout problem,’’ Can.
[90] C. Mao, ‘‘Harmony search-based test data generation for branch coverage J. Civil Eng., vol. 37, no. 12, pp. 1560–1571, 2010.
in software structural testing,’’ Neural Comput. Appl., vol. 25, no. 1, [114] R. Mikaeil, S. S. Haghshenas, Y. Shirvand, M. V. Hasanluy, and
pp. 199–216, 2014. V. Roshanaei, ‘‘Risk assessment of geological hazards in a tunnel-
[91] S. A. Abed, S. Tiun, and N. Omar, ‘‘Harmony search algorithm for word ing project using harmony search algorithm (case study: Ardabil-
sense disambiguation,’’ PLoS ONE, vol. 10, no. 9, p. e0136614, 2015. Mianeh railway tunnel),’’ Civil Eng. J., vol. 2, no. 10, pp. 546–554,
[92] E. Cuevas, N. Ortega-Sánchez, D. Zaldivar, and M. Pérez-Cisneros, ‘‘Cir- 2016.
cle detection by harmony search optimization,’’ J. Intell. Robot. Syst., [115] J. Park, K.-H. Lee, J. Park, H. Choi, and I.-M. Lee, ‘‘Predicting anomalous
vol. 66, no. 3, pp. 359–376, 2012. zone ahead of tunnel face utilizing electrical resistivity: I. Algorithm
[93] A. Vasebi, M. Fesanghary, and S. M. T. Bathaee, ‘‘Combined heat and and measuring system development,’’ Tunnelling Underground Space
power economic dispatch by harmony search algorithm,’’ Int. J. Elect. Technol., vol. 60, pp. 141–150, Nov. 2016.
Power Energy Syst., vol. 29, no. 3, pp. 713–719, 2007. [116] A. H. Kayhan, K. A. Korkmaz, and A. Irfanoglu, ‘‘Selecting and scaling
[94] E. Khorram and M. Jaberipour, ‘‘Harmony search algorithm for solving real ground motion records using harmony search algorithm,’’ Soil Dyn.
combined heat and power economic dispatch problems,’’ Energy Convers. Earthquake Eng., vol. 31, no. 7, pp. 941–953, 2011.
Manage., vol. 52, no. 2, pp. 1550–1554, 2011. [117] H. Ceylan, H. Ceylan, S. Haldenbilen, and O. Baskan, ‘‘Transport
[95] A. H. Khazali and M. Kalantar, ‘‘Optimal reactive power dispatch based energy modeling with meta-heuristic harmony search algorithm, an
on harmony search algorithm,’’ Int. J. Elect. Power Energy Syst., vol. 33, application to Turkey,’’ Energy Policy, vol. 36, no. 7, pp. 2527–2535,
no. 3, pp. 684–692, 2011. 2008.

14244 VOLUME 7, 2019

A. A. Al-Omoush et al.: Comprehensive Review of the Development of the HS Algorithm and Its Applications

[118] S. Mun and Z. W. Geem, ‘‘Determination of viscoelastic and damage ABDULRAHMAN A. ALSEWARI was born
properties of hot mix asphalt concrete using a harmony search algorithm,’’ in Sana’a, Yemen, in 1980. He received the
Mech. Mater., vol. 41, no. 3, pp. 339–353, 2009. B.Eng. degree in computer engineering from the
[119] P. Yadav, R. Kumar, S. K. Panda, and C. S. Chang, ‘‘An improved har- Military College of Engineering, Baghdad, Iraq,
mony search algorithm for optimal scheduling of the diesel generators in in 2002, and the M.Sc. degree in electronic system
oil rig platforms,’’ Energy Convers. Manage., vol. 52, no. 2, pp. 893–902, design engineering and the Ph.D. degree in soft-
2011. ware engineering from Universiti Sains Malaysia,
[120] S. Ryu, A. S. Duggal, C. N. Heyl, and Z. W. Geem, ‘‘Cost-optimized
in 2008 and 2012, respectively. From 2012 to
FPSO mooring design via harmony search,’’ J. Offshore Mech. Arctic
2013, he was an Assistant Professor and the
Eng., vol. 138, no. 6, p. 061303, 2016.
[121] O. Zarei, M. Fesanghary, B. Farshi, R. Jalili Saffar, and M. R. Razfar, Head of the Information Technology Department,
‘‘Optimization of multi-pass face-milling via harmony search algorithm,’’ Engineering College, Darussalam International University for Science and
J. Mater. Process. Technol., vol. 209, no. 3, pp. 2386–2392, 2009. Technology, Yemen. Since 2013, he has been a Senior Lecturer (Assistant
[122] M. Huang, H.-Y. Dong, X.-W. Wang, B.-L. Zheng, and W. H. Ip, ‘‘Guided Professor) and a Researcher with the Software Engineering Department, Fac-
variable neighborhood harmony search for integrated charge planning ulty of Computer Systems and Software Engineering, Universiti Malaysia
in primary steelmaking processes,’’ in Proc. 1st ACM/SIGEVO Summit Pahang. He has authored or co-authored one book, more than 30 articles, and
Genet. Evol. Comput., 2009, pp. 231–238. more than 10 inventions. His research interests include software engineering,
[123] A. M. Mohsen, A. T. Khader, and D. Ramachandram, ‘‘Hsrnafold: software testing, optimization algorithms, artificial intelligence, soft com-
A harmony search algorithm for rna secondary structure prediction based puting, embedded systems, and image processing. He is a Senior Member of
on minimum free energy,’’ in Proc. Int. Conf. Innov. Inf. Technol. (IIT), the IEEE.
Dec. 2008, pp. 11–15.
[124] A. M. Mohsen, A. T. Khader, and A. Ghallab, ‘‘The effect of using
different thermodynamic models with harmony search algorithm in the
accuracy of RNA secondary structure prediction,’’ in Proc. Int. Conf. Soft
Comput. Pattern Recognit. (SOCPAR), Dec. 2009, pp. 505–510.
[125] J. Amor, E. Alexandre, R. Gil, L. Álvarez, and E. Huerta, ‘‘Music-
inspired harmony search algorithm applied to feature selection for sound
classification in hearing AIDS,’’ in Proc. Audio Eng. Soc. Conv., 2008,
p. 124.
[126] K. S. Hickmann, ‘‘Forecasting the 2013-2014 influenza season using
Wikipedia,’’ PLoS Comput. Biol., vol. 11, no. 5, p. e1004239, 2015.
HAMMOUDEH S. ALAMRI received the B.Sc.
[127] S. Taghipour, P. Zarrineh, M. Ganjtabesh, and A. Nowzari-Dalini,
‘‘Improving protein complex prediction by reconstructing a high- degree in computer engineering from Yarmouk
confidence protein-protein interaction network of Escherichia coli from University, in 2008, and the M.Sc. degree from
different physical interaction data sources,’’ BMC Bioinf., vol. 18, no. 1, University Malaysia Pahang, in 2016, where he is
p. 10, 2017. currently pursuing the Ph.D. degree in computer
[128] J. Dongardive, A. Patil, A. Bir, S. Jamkhedkar, and S. Abraham, ‘‘Finding science with the Faculty of Computer Systems and
motifs using harmony search,’’ in Proc. Int. Symp. Biocomput., 2010, Software Engineering. His main research interests
p. 41. include optimization algorithms, the IoT, and com-
[129] S. Tuo, J. Zhang, X. Yuan, Z. He, Y. Liu, and Z. Liu, ‘‘Niche harmony putational intelligence.
search algorithm for detecting complex disease associated high-order
SNP combinations,’’ Sci. Rep., vol. 7, no. 1, 2017, Art. no. 11529.
[130] S. Mun and Z. W. Geem, ‘‘Determination of individual sound power levels
of noise sources using a harmony search algorithm,’’ Int. J. Ind. Ergonom.,
vol. 39, no. 3, pp. 366–370, 2009.
[131] A. Panchal, ‘‘Harmony search in therapeutic medical physics,’’ in Music-
Inspired Harmony Search Algorithm. Berlin, Germany: Springer, 2009,
pp. 189–203.
[132] O. M. Alia, R. Mandava, D. Ramachandram, and M. E. Aziz, ‘‘Harmony
search-based cluster initialization for fuzzy c-means segmentation of MR
images,’’ in Proc. IEEE Region 10 Conf. (TENCON), Jan. 2009, pp. 1–6.
[133] S. Ngonkham and P. Buasri, ‘‘Harmony search algorithm to improve cost
reduction in power generation system integrating large scale wind energy
conversion system,’’ in Proc. World Non-Grid-Connected Wind Power
Energy Conf. (WNWEC), Sep. 2009, pp. 1–5.
[134] F. Erdal E. Doǧan, and M. P. Saka, ‘‘Optimum design of cellular beams
KAMAL Z. ZAMLI received the degree in elec-
using harmony search and particle swarm optimizers,’’ J. Construct. Steel
trical engineering from Worcester Polytechnic
Res., vol. 67, no. 3, pp. 237–247, 2011.
Institute, USA, in 1992, the M.Sc. degree in
real-time software engineering from Universiti
ALA’A A. AL-OMOUSH received the B.Sc. degree Teknologi Malaysia, in 2000, and the Ph.D. degree
in computer science from Al Albayt University, in software engineering from the University of
in 2010, and the M.Sc. degree from University Newcastle upon Tyne, U.K., in 2003. He is cur-
Putra Malaysia, in 2013. He is currently pursuing rently a Professor and the Dean of the Faculty
the Ph.D. degree in computer science with the of Computer Systems and Software Engineering,
Faculty of Computer Systems and Software Engi- Universiti Malaysia Pahang. He has authored or
neering, Universiti Malaysia Pahang. His main co-authored one book, more than 30 articles, and more than 350 papers in
research interests include artificial intelligence and journals and conferences worldwide mainly in the areas of (combinatorial
computational intelligence. t-way) software testing and search-based software engineering.

VOLUME 7, 2019 14245