Expert Systems With Applications 239 (2024) 122413

Contents lists available at ScienceDirect

Expert Systems With Applications

journal homepage: www.elsevier.com/locate/eswa

Walrus optimizer: A novel nature-inspired metaheuristic algorithm

Muxuan Han a, Zunfeng Du a, *, Kum Fai Yuen b, Haitao Zhu a, Yancang Li c, Qiuyu Yuan a
a
State Key Laboratory of Hydraulic Engineering Intelligent Construction and Operation, School of Civil Engineering, Tianjin University, Tianjin 300354, China
b
School of Civil and Environmental Engineering, Nanyang Technological University, 639798, Singapore
c
College of Civil Engineering, Hebei University of Engineering, Handan 056038, China

A R T I C L E I N F O A B S T R A C T

Keywords: Metaheuristic algorithms are intelligent optimization approaches that lead the searching procedure through
Walrus Optimizer (WO) utilizing exploitation and exploration. The increasing complexity of real-world optimization problem has
Metaheuristic algorithm prompted the development of more metaheuristic algorithms. Hence, this work proposes a novel swarm intel­
Swarm intelligence
ligence algorithm, Walrus optimizer (WO). It is inspired by the behaviors of walruses that choose to migrate,
Exploration
Exploitation
breed, roost, feed, gather and escape by receiving key signals (danger signals and safety signals). To test the
capability of the proposed algorithm, 23 standard functions and the benchmark suite from the IEEE (Institute of
Electrical and Electronics Engineers) Congress on Evolutionary Computation (CEC) 2021 are used. In addition, to
evaluate the practicability of the proposed algorithm to solve various real-world optimization problems, 6
standard classical engineering optimization problems are examined and compared. For statistical purposes, 100
independent optimization runs are conducted to determine the statistical measurements, including the mean,
standard deviation, and the computation time of the program, by considering a predefined stopping criterion.
Some well-known statistical analyses are also used for comparative purposes, including the Friedman and Wil­
coxon analysis. The results demonstrate that the proposed algorithm can provide special stability features and
very competitive performance in dealing with high-dimensional benchmarks and real-world problems. The
proposal of WO promotes the continuous development and application expansion of artificial intelligence, im­
proves the efficiency of optimization calculation, and provides powerful tools for solving complex problems in
the real world. The source code of WO is publicly availabe at https://ww2.mathworks.cn/matlabcentral/fileexch
ange/154702-walrus-optimizer-wo.

1. Introduction deterministic and have clear requirements on mathematical models and

constraints. The typical traditional optimization methods include Linear
The general formulation of an optimization problem is to select a set Programming (LP), Non-Linear Programming (NLP), Mixed-Integer
of parameters (variables) and make the design specifications (objec­ Programming (MIP), and Dynamic Programming (DP). Although re­
tives) reach the optimal maximum or minimum value under a series of searchers have proposed many different methods, the following prob­
relevant constraints (limitations). Therefore, optimization problems can lems still exist:
usually be expressed in the form of mathematical programming. Solving
optimization problems is the norm in almost all disciplines of science (1) Most of the traditional optimization methods have a high
and engineering, and the need for more robust solutions is ever dependence on the mathematical model. If the objective function
increasing. That means, we need reasonable optimization methods that and constraint conditions cannot meet the conditions of first-
can fit the intricate nature of such up-to-date scientific and engineering order and second-order differentiability, it can fail to converge.
challenges. (2) The selection of initial points will affect the entire solution pro­
Mathematical optimization techniques used to be the only tools for cess of the problem. If the error of the selection of initial points is
optimizing problems before the proposal of heuristic optimization large, it will lead to low accuracy of the calculation results.
techniques. Mathematical optimization methods are mostly

* Corresponding author.
E-mail addresses: [email protected] (M. Han), [email protected] (Z. Du), [email protected] (K.F. Yuen), [email protected] (H. Zhu), liyancang@hebeu.
edu.cn (Y. Li), [email protected] (Q. Yuan).

https://doi.org/10.1016/j.eswa.2023.122413
Received 20 May 2023; Received in revised form 17 October 2023; Accepted 29 October 2023
Available online 3 November 2023
0957-4174/© 2023 Elsevier Ltd. All rights reserved.
M. Han et al. Expert Systems With Applications 239 (2024) 122413

(3) Actual optimization problems are usually characterized by com­ selection, design, and adjustment of metaheuristic algorithms need to
plex structure and large scale. If traditional optimization methods consider the specific requirements and characteristics of the problem to
are used for optimization, problems such as complex computa­ achieve the best performance.
tion, difficult to find a solution and long solution time will arise. Combining the above two theories, it is necessary to develop a novel
metaheuristic algorithm to solve combinatorial optimization problems.
Due to the inability of accurate optimization methods to solve The specific motivations are listed below.
complex and multidimensional problems, approximation algorithms
have been proposed to solve such problems. Approximation algorithms (1) Improving the efficiency of solving complex problems: meta­
are a new approach for efficiently solving optimization problems and are heuristic algorithms are usually used to solve complex optimi­
divided into two categories: heuristic and metaheuristic algorithms zation problems, such as travel salesman problems, scheduling
(Abdollahzadeh et al., 2021; Ong et al., 2021; X. S. Yang, 2010). problems, path planning or even multi-objective optimization
In mathematical programming, a heuristic algorithm is a procedure problems. The metaheuristic algorithms introduce a series of
that determines near-optimal solutions to an optimization problem. intelligent heuristic strategies and techniques to improve the ef­
However, this is achieved by trading optimality, completeness, accu­ ficiency of solving complex problems and reduce the computa­
racy, or precision for speed. Heuristic algorithms construct feasible so­ tional cost. Novel metaheuristic algorithms use a combination of
lutions in a certain number of steps for a specific problem feasible set, i. local and global search, which balances the trade-off between fast
e., reflected as problem dependent (Kuo & Prasad, 2000; Shin et al., convergence and avoiding falling into local optimal solutions.
2008). A feasible solution for each instance of the optimization problem (2) Dealing with large-scale data: With the advent of the big data era,
to be solved is given at an acceptable cost (in terms of computational many problems require dealing with large-scale datasets. Meta­
time, computational space, etc.), and the degree of deviation of the heuristic algorithms can be used to accelerate the analysis and
feasible solution from the optimal solution is generally not predictable in processing of these data. It can help decision makers get useful
advance. However, since they tend to be too greedy, they usually fall information faster and improve the feasibility of problem solving.
into a local optimum and thus usually fail to obtain a globally optimal (3) Adapting to problem uncertainty: Some problems have an
solution. element of uncertainty, such as randomness or incomplete in­
Metaheuristic algorithms are strategies that guide the search process. formation. Through diverse strategies such as fuzzy logic, robust
The goal is to efficiently explore the search space to find near–optimal optimization, stochastic probabilistic modeling, and reinforce­
solutions. Techniques which constitute metaheuristic algorithms range ment learning, metaheuristic algorithms can be designed to be
from simple local search procedures to complex learning processes. They more flexible, allowing them to perform well in problems with
do not exploit any specificity of the problem and thus can be used as a uncertainty factors.
black box, i.e., reflected as problem independent. In general, they are (4) Theoretical research: theoretical research on metaheuristic al­
not greedy. In fact, they may even accept a temporary deterioration of gorithms is an important motivation. By studying new meta­
the solution in a specific problem, which allows them to explore the heuristic algorithms, the theory of computation can be advanced
solution space more thoroughly, leading to a hopefully better solution to gain insight into the complexity and solvability of problems.
(sometimes coinciding with the global optimum). (5) Innovation and competitive advantage: In scientific research and
Another difference between heuristic and metaheuristic algorithms industrial applications, proposing new metaheuristic algorithms
is the presence of a “random factor”. For the same problem, given an can lead to innovation and competitive advantage. This can
input, the steps performed by the heuristic algorithm are fixed, and so is motivate researchers to invest more resources in developing new
the output. This is not the case with metaheuristic algorithms, which algorithms.
include a “random factor”. In the case of a Genetic Algorithm (GA)
(Holland, 1992), if the same initial population (with the same inputs) is Therefore, Walrus Optimizer (WO) is proposed in this paper with a
run twice (100 generations each time), the results will probably be simple structure, special stability features and very competitive perfor­
different. This is because the objects of the crossover and mutation op­ mance to realize the solutions of both constrained and unconstrained
erations are chosen randomly. The success of GA depends largely on the problems more effectively. The innovation is that the “vigilantes” in a
random component of the algorithm. The overall fitness gets improved walrus herd are the determining factor influencing the direction of the
over the process of generations as the best individuals are more likely to herd. The “Danger signal” and “Safety signal” issued by the “vigilantes”
select and reproduce than the worst individuals. play a key role in the execution process of WO. In WO, the
Complexity theory (Brown et al., 2016; Ladyman et al., 2013) and No “Danger signal” is used to determine whether the WO performs the
Free lunch (NFL) theorem (Ho & Pepyne, 2002) are two important exploration phase or the exploitation phase. When the “Danger signal”
concepts used to elucidate the limitations of metaheuristic algorithms. meets certain conditions, the walrus herd migrates to a new domain
Complexity theory focuses on the relationship between problem diffi­ within the solution space, which is the exploration phase in the early
culty and computational resources. The theory categorizes problems stage of the algorithm. On the contrary, the walrus herd reproduces,
into P problems (polynomial time solvable) and NP problems (non- which is the exploitation phase in the late stage of the algorithm. The
deterministic polynomial time solvable). The theory suggests that met­ “Safety signal” plays a key role in the exploitation phase, which in­
aheuristic algorithms are usually not guaranteed to find a globally fluences whether a walrus chooses the roosting behavior or foraging.
optimal solution, which is particularly evident in the case of NP-hard Among them, male walruses, female walruses, and juvenile walruses
problems (Sörensen, 2015). They usually stop at locally optimal or interact with each other in their roosting behaviors to move the popu­
suboptimal solutions. This reflects the limited nature of meta-heuristic lation in a direction conducive to survival; foraging behaviors include
algorithms, as they cannot break the fundamental limitations of the typical phenomena of gathering and fleeing, which are controlled by
complexity theory. The NFL theorem states that there is no algorithm for “Danger signal”. The walrus herd can avoid capture or death by preda­
solving all optimization problems, which means that the performance of tors (falling into the local optimum) and achieve population growth
different algorithms is not equal when given the prior knowledge (al­ (searching for the global optimum) in an orderly policing environment.
gorithm parameters, convergence criteria) of a meta-heuristic algorithm In the remainder of the paper, Section 2 deals with the related works.
for solving a particular problem (Xu et al., 2023). Concisely, there is no Section 3 shows the model of the proposed WO. In Section 4, the pro­
universal optimization procedure that works perfectly for all optimiza­ posed WO is evaluated by 23 benchmark functions, CEC 2021 bench­
tion problems and thus, the continuous flourish of the diversity of mark suit, 0-1 knapsack problems, and six engineering design problems.
optimization algorithm is encouraged. In practical applications, the Finally, the conclusions are summarized in Section 5.

2
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Table 1
Summary of the latest metaheuristic algorithms.
Category Algorithm Year Ref. Inspiration

SI African Vultures Optimization 2021 (Abdollahzadeh et al., 2021) The navigation behaviors of African vultures
Algorithm (AVOA)
Golden Eagle Optimizer (GEO) 2021 (Mohammadi-Balani et al., The spiral flight trajectory of golden eagles
2021)
Red Fox Optimization (RFO) 2021 (Połap & Woźniak, 2021) The behaviors of red foxes using various tricks to distract prey
Aquila Optimizer (AO) 2021 (Abualigah et al., 2021) The behaviors of Aquila during prey capture
Tuna Swarm Optimization (TSO) 2021 (Xie et al., 2021) The cooperative foraging behavior of tuna swarm
Wild Horse Optimizer (WHO) 2021 (Naruei & Keynia, 2021b) The social life behaviors of wild horses
Dingo Optimization Algorithm (DOA) 2021 (Peraza-Vázquez et al., 2021) The hunting strategies of dingoes which are attacking by persecution, grouping
tactics, and scavenging behavior
Coot Optimization Algorithm (COOT) 2021 (Naruei & Keynia, 2021a) The behavior of Coots on the water surface
Wild Horse Optimizer (WHO) 2021 (Naruei & Keynia, 2021b) The social life behaviors of wild horses
Chameleon Swarm Algorithm (CSA) 2021 (M. S. Braik, 2021) The navigating and hunting behaviors of chameleons
Capuchin Search Algorithm (CapSA) 2021 (M. Braik et al., 2021) The dynamic behavior of capuchin monkeys jumping, swinging, and climbing
Northern Goshawk Optimization 2021 (Dehghani et al., 2021) The behavior of northern goshawk during prey hunting
(NGO)
Hunter-Prey Optimization (HPO) 2022 (Naruei et al., 2022) The behavior of predator and prey
Archerfish Hunting Optimizer (AHO) 2022 (Zitouni et al., 2022) The shooting and jumping behaviors of the archerfish for hunting aerial insects
Honey Badger Algorithm (HBA) 2022 (Hashim et al., 2022) The intelligent foraging behavior of honey badger
Artificial Rabbits Optimization (ARO) 2022 (L. Wang et al., 2022) The survival strategies of rabbits
Beluga Whale Optimization (BWO) 2022 (Zhong et al., 2022) The behaviors of pair swim, prey, and whale fall
Dwarf Mongoose Optimization 2022 (Agushaka et al., 2022) The foraging behavior of the dwarf mongoose
Algorithm (DMO)
Golden Jackal Optimization (GJO) 2022 (Chopra & Mohsin Ansari, The collaborative hunting behavior of the golden jackals
2022)
Sand Cat Swarm Optimization (SCSO) 2022 (Seyyedabbasi & Kiani, 2022) The sand cat incredible ability to dig for prey
Snake Optimizer (SO) 2022 (Hashim & Hussien, 2022) The special mating behavior of snakes
White Shark Optimizer (WSO) 2022 (M. Braik et al., 2022) The behaviors of white sharks, including exceptional senses of hearing and
smell
Reptile Search Algorithm (RSA) 2022 (Abualigah et al., 2022) The hunting behaviors of Crocodiles
Coati Optimization Algorithm (COA) 2022 (Dehghani et al., 2023) The behavior of coati when attacking and hunting iguanas
Nutcracker Optimizer (NO) 2023 (Abdel-Basset et al., 2023) The search, cache, and recovery behaviors of the nutcracker
Bedbug Meta-Heuristic Algorithm 2023 (Rezvani et al., 2023) The static and dynamic swarming behaviors of bedbugs
(BMHA)
Willow Catkin Optimization (WCO) 2023 (J.-S. Pan et al., 2023) Willow trees’ process of seed dispersal
Osprey Optimization Algorithm (OOA) 2023 (Dehghani & Trojovský, 2023) The hunting strategy of ospreys
EA Coronavirus Herd Immunity 2021 (Al-Betar et al., 2021) The herd immunity concept to tackle coronavirus pandemic (COVID-19)
Optimization (CHIO)
Ebola Optimization Search Algorithm 2022 (Olaide et al., 2022) The propagation method of the Ebola virus disease
(EOSA)
PhA Solar System Algorithm (SSA) 2021 (Zitouni et al., 2021) The orbiting behavior of some objects found in the solar system
Crystal Structure Algorithm (CryStAl) 2021 (Talatahari et al., 2021) The principles underlying the formation of crystal structures
Arithmetic Optimization Algorithm 2021 (Abualigah et al., 2021) The arithmetic operators in mathematics
(AOA)
Atomic Orbital Search (AOS) 2021 (Azizi, 2021) The principles of quantum mechanics and the atomic model
Volcano Eruption Algorithm (VEA) 2021 (Hosseini et al., 2021) The nature of volcano eruption
RUNge Kutta Optimizer (RUN) 2021 (Ahmadianfar et al., 2021) The Runge Kutta (RK) method
Lichtenberg Algorithm (LA) 2021 (Pereira et al., 2021) The Lichtenberg figures patterns
Chernobyl Disaster Optimizer (CDO) 2023 (Shehadeh, 2023) The nuclear reactor core explosion of Chernobyl
Energy Valley Optimizer (EVO) 2023 (Azizi et al., 2023) The physics principles regarding stability and different modes of particle decay
Human Cooperation Search Algorithm (CSA) 2021 (Feng et al., 2021) The team cooperation behaviors in modern enterprise
based Stock Exchange Trading Optimization 2021 (Emami, 2022b) The behavior of traders in the stock market
(SETO)
Social Network Search (SNS) 2021 (Talatahari et al., 2021) The behavior of users expressing their opinions
Giza Pyramids Construction (GPC) 2021 (Harifi et al., 2021) The ancient past
Hunger Games Search (HGS) 2021 (Y. Yang et al., 2021) Animal hunger drives activity and behavior design
Human Felicity Algorithm (HFA) 2022 (Verij kazemi & Fazeli The efforts of human society to become felicity
Veysari, 2022)
Ali Baba and the Forty Thieves (AFT) 2022 (M. Braik et al., 2022) The strategies pursued by the forty thieves in the search for Ali Baba
Alpine Skiing Optimization (ASO) 2022 (Yuan et al., 2022) The behavior of skiers competing for the championship
Anti-Coronavirus Optimization (ACVO) 2022 (Emami, 2022a) The behavior of people to slow the spread of COVID-19
Musical Chairs Algorithm (MCA) 2023 (Eltamaly & Rabie, 2023) The musical chairs game
Group Learning Algorithm (GLA) 2023 (Rahman, 2023) The way individuals inside a group affect each other
Mountaineering Team-Based 2023 (Faridmehr et al., 2023) A cooperative human phenomenon
Optimization (MTBO)

2. Related works (1) Swarm intelligence (SI)

Metaheuristics can be classified according to various characteristics SI is a general term for a class of intelligent groups with self-
(Chica et al., 2017): nature-inspired vs. not nature-inspired; deterministic organized behavior. SI can be seen as an intelligent behavior formed
vs. stochastic; population-based vs. single-solution-based search; and by interactions between individuals and individuals, individuals, and
iterative vs. greedy. Moreover, metaheuristic algorithms can be catego­ environment. Individuals in a group all follow a simple code of conduct
rized into the following four groups based on the source of inspiration. with no unified central control between groups. The interaction between

3
M. Han et al. Expert Systems With Applications 239 (2024) 122413

individuals is ultimately manifested as the intelligence of the whole algorithm merging the SEO and the particle swarm algorithm (PSO) and
group. applied it to Structural Health Monitoring (SHM) of frame structures.
Hashim et al. (Hashim et al., 2022) proposed the Honey Badger Al­ In this section, new metaheuristic algorithms from 2021 to 2023 are
gorithm (HBA) inspired by the intelligent foraging behavior of honey summarized, but is not limited to those presented in Table 1. In addition
badger. HBA has been applied in sliding mode controller (T. Wang et al., to the four major categories of metaheuristic algorithms mentioned
2023) and optimal power flow (Akdağ, 2022). Fathollahi-Fard et al. above, many scholars have designed efficient optimization algorithms
(Fathollahi-Fard et al., 2020) proposed the Red Deer Algorithm (RDA) for special problems. It may contain the basic ideas of one or more al­
based on the unusual mating behavior of Scottish red deer in a breading gorithms, and the mixed algorithms play an irreplaceable role in specific
season. RDA has been applied to solve optimal location problem of problem solving. For example, Fathollahi-Fard et al. (Fathollahi-Fard
distributed generation (Lakshmi et al., 2023), Sugarcane Supply Chain et al., 2023) proposed an adaptive large neighborhood search algorithm
Network (SSCN) (Chouhan et al., 2021a) and data clustering (Mogha­ based on heuristics and reconstruction for the generalized quadratic
dam & Ahmadi, 2023). The Keshtel Algorithm (KA) (Hajiaghaei-Keshteli assignment problem.
& Aminnayeri, 2013) derives its inspiration from the foraging behavior The study of novel metaheuristic algorithms has been one of the
of the Keshtel, a dabbling duck. With its emergence, KA has been widely important research directions in the field of computer science and arti­
used in various fields, including integrated scheduling (Hajiaghaei- ficial intelligence, but there are still many research gaps that need to be
Keshteli & Aminnayeri, 2014), supply chain networks design (Chouhan further explored and solved:
et al., 2022; Mosallanezhad et al., 2021, 2023; Zahedi et al., 2021), and
allocation-routing optimization models (Hashemi-Amiri et al., 2023). (1) Data imbalance: In many practical problems, data sets are often
imbalanced. Some of these categories have a much larger sample
(2) Evolutionary algorithms (EA) size than others. Novel algorithms need to deal with this data
imbalance to improve the performance of the model.
EA is inspired by the biological evolution of nature and have self- (2) Small-sample learning: in many domains, only a very limited
organizing, self-adaptive and self-learning properties. Biological evolu­ amount of labeled data is available for training models. There­
tion is achieved through reproduction, variation, competition, and se­ fore, studying algorithms for small sample learning is a chal­
lection. The EA can solve the optimization problem mainly through lenging gap.
selection, recombination, and mutation. (3) Interdisciplinary research: the study of intelligent algorithms
Al-Betar et al. (Al-Betar et al., 2021) proposed the Coronavirus Herd involves many fields such as computer science, mathematics,
Immunity Optimizer (CHIO), which is originated from the herd immu­ statistics, neuroscience and so on. Exploring the intersection be­
nity concept as a way to tackle coronavirus pandemic (COVID-19). tween these fields for deeper understanding and innovation is a
Kumar et al. (Kumar et al., 2022) proposed the Multi-Objective Coro­ promising direction.
navirus Herd Immunity Optimizer (MOCHIO) and applied it to the (4) Practical applications: despite the many theoretical potentials of
optimization of brushless direct current (BLDC) motor design. Abu intelligent algorithms, there are still challenges in applying them
Doush et al. (Abu Doush et al., 2023) applied the CHIO to the training to solve practical problems. Researchers can focus on how to
process of multilayer perceptron neural network (MLP) to find its effectively apply novel intelligent algorithms to domains such as
optimal control parameters, thus empowering their classification healthcare, finance, and energy management.
accuracy.
These research gaps represent some of the current challenges and
(3) Physics-based algorithms (PhA) opportunities in the field of intelligent algorithms. Addressing these
challenges will help to further advance the development of Artificial
PhA is stimulated from the fundamental physical laws of the universe Intelligence (AI) technology to better serve the needs of society and
and usually reflects the main rules of physical processes in the in­ science.
terconnections between search agents in the algorithm execution.
The mathematical development of the Atomic Orbital Search (AOS) 3. Walrus Optimizer (WO)
is based on principles of quantum mechanics focusing on the act of
electrons around the nucleus of an atom (Azizi, 2021). Talatahari et al. The following is about the inspiration, mathematical model, pro­
(Talatahari et al., 2021) proposed the Crystal Structure algorithm cedure and complexity for WO.
(CryStAl). This method is chiefly inspired by the principles underlying
the formation of crystal structures from the addition of the basis to the 3.1. Biological fundamentals
lattice points. Shehadeh (Shehadeh, 2023) was inspired by the Cher­
nobyl nuclear reactor core explosion to propose the Chernobyl Disaster The walrus is the largest mammal in the ocean besides whales.
Optimizer (CDO). Walruses live mainly in temperate waters in or near the Arctic. Walruses
are gregarious animals, leading an amphibious life. Groups of walruses
(4) Human-based algorithms range from dozens to hundreds to thousands of individuals. Its body is
cylindrical, stout, and obese, with a flattened head and blunt muzzle
The last category of metaheuristic algorithms presented is human- end. Around the upper lip there are about 400 long and hard whiskers
based algorithms which contain algorithms inspired by human beings, with blood vessels and nerves, leading to a sharp sense of touch. The
including physical and non-physical activities such as thinking and so­ most unique feature of the walrus is that pair of white, well-developed
cial behavior. upper canine teeth that keep growing throughout its life, forming
Fathollahi-Fard et al. (Fathollahi-Fard et al., 2018) proposed the tusks. The tusks can be used for self-defense, digging in the mud and
Social Engineering Optimizer (SEO). The basis of SEO depends on how sand for food such as clams, shrimp, and crabs, or to support the body
an attacker attacks to a defender by four different social engineering when climbing on the ice (Gotfredsen et al., 2018).
techniques, and it precisely completes the reinforcement and the Walruses look bulky with their massive bodies, but flexible in the
diversification phases. SEO has been widely used in supply chain water. Among that large number of marine animals, walruses are the
network (Mousavi et al., 2021; Salehi-Amiri et al., 2021; Zahedi et al., best divers. They can dive in the water for 20 min and to a depth of 500
2021) and mixed-integer linear programming modelling (Chouhan m. Walruses can stay underwater for up to 2 h after diving to the bottom,
et al., 2021b). Alkayem et al. (Alkayem et al., 2022) proposed a new and once need fresh air, they can surface within 3 min.

4
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Fig. 1. Herding behavior and population distribution of walruses. (a) Herding behavior (b) Population distribution.

Walruses have a strong sense of community as shown below: variables, rand is a uniform random vector in the range 0 to 1.
Walruses are agents that perform the optimization process. Their
• When the breeding season begins, walruses establish their territories positions are continuously updated with iterations.
on the beach (Ray et al., 2016). The best positions are occupied by ⎡ ⎤
the strongest males. The area of the territory varies according to the ⎢
X1,1 X1,2 ⋯X1,d
⎥
number of females occupied by the male. ⎢ X2,1 X2,2 ⋯X2,d ⎥
⎢ ⎥
• Walruses are accustomed to living in deep water areas of the ocean X=⎢
⎢ ⎥
⎥ (2)
⎢ ⋮⋮⋮⋮ ⎥
where sunlight cannot reach. Like bats and dolphins, the walruses ⎢ ⎥
⎢ ⎥
are lack of idiosyncratic vision; They rely on sound localization for ⎣ ⋮⋮⋮⋮ ⎦
foraging and share food information by communicating with peers. Xn,1 Xn,2 ⋯Xn,d
n×d
• Walruses have strong social habits when they encounter killer whales
in the water, they will adopt a collective defense strategy to defend where, n is the population size, d is the dimension of design variables.
themselves and aid their injured counterparts. The experience of the The fitness values corresponding to all search agents are stored as:
long struggle for survival prevents the walrus from letting down its ⎡ ⎤
guard (Jay et al., 1998). At that point, two walruses act as guards. ⎢
(f1,1 f1,2 ⋯f1,d )
⎥
They will go to help once their kind is injured. ⎢
⎢ (f2,1 f2,2 ⋯f2,d ) ⎥
⎥
⎢ ⎥
⎢ ⎥
F=⎢ ⋮⋮⋮⋮ ⎥ (3)
Inspired from the behaviors of walrus in migrating, breeding, ⎢ ⎥
⎢ ⎥
roosting and foraging (as shown in Fig. 1), we propose a new meta­ ⎢ ⋮⋮⋮⋮ ⎥
⎣ ⎦
heuristic algorithm, WO, for the first time. Two assumptions need to be (fn,1 fn,2 ⋯fn,d )
clarified here: n×d

Walrus populations are divided into adults and juveniles, which ac­
(1) Walrus populations judge population behavior by danger and count for 90 % and 10 % of the population, respectively. Among adult
safety signals. walruses, the male to female ratio is 1:1.
(2) Behavioral and role divisions in walrus populations are modeled
in the walrus algorithm. Specifically, the walrus algorithm as­ 3.2.2. Danger signals and safety signals
sumes social structures and interactions between male, female, Walruses are very vigilant, both in foraging and roosting. There will
and juvenile walruses. be 1 to 2 walruses as guards patrolling around, and danger signals will
be sent out immediately once unexpected situations are found. The
In the WO mathematical model, the search space is the range over danger signal and safety signal in WO are defined as follows:
which the algorithm tries to find the best solution. It is usually a
Danger signal = A*R (4)
multidimensional space consisting of the decision variables of the
problem. The solution space contains all potential solutions; A solution
α = 1 − t/T (5)
is the location of a walrus in the search space. It represents a potential
solution to the problem. The solution can be a vector, where each A=2×α (6)
component corresponds to the value of a decision variable; The decision
variable is determined by the problem. WO tries to find the best solution R = 2 × r1 − 1 (7)
in the search space, i.e., the solution that optimizes the objective func­
tion of the problem. where, A and R are danger factors, α decreases from 1 first to 0 with the
number of iterations t, and T is the maximum iteration.
The safety signal corresponding to the danger signal in WO is defined
3.2. Mathematical model and algorithm
as follows:
3.2.1. Initialization Safety signal = r2 (8)
In WO, the optimization process starts with a set of randomly
generated candidate solutions (X). where, r1 and r2 are random numbers lie in the range of (0, 1).

X = LB + rand(UB − LB) (1) 3.2.3. Migration (exploration)

When risk factors are too high, walrus herds will migrate to areas
where, LB and UB are the lower and upper boundary of the problem

5
M. Han et al. Expert Systems With Applications 239 (2024) 122413

more suitable for population survival. In this phase, the walrus position random number of (0,1), O is the reference safety position, LF is a vector
is updated as follows: of random numbers based on Lévy distribution representing Lévy
movement.
Xi,jt+1 = Xi,jt + Migration step (9)
́ x
Levy(a) = 0.05 × (15)
(10)
1
Migration step = (Xmt − Xnt ) • β • r3 2 |y|a

1 where x and y are two normally distributed variables, x N(0, σ2x ), y N(0,
β = 1− (11)
1 + exp(−
t− T2
T
× 10) σ 2y ).
⎡ ( ) ⎤α1
t+1
where, Xi,j is the new position for the i th walrus on the j th dimension, Γ(1 + α )sin πα
⎢ 2 ⎥
t
Xi,j is the current position of the i th walrus on the j th dimension, σx = ⎢ ⎥
⎣ ( ) (α− 1) ⎦ , σy = 1, α = 1.5 (16)
Migration step is the step size of walrus movement, two vigilantes are
1+α
Γ 2 α2 2
randomly selected from the population, the positions of the vigilantes
t
correspond to Xm and Xnt , β is the migration step control factor, which where, σ x and σ y are the standard deviations, Γ(x) = (x + 1)!.
varies with iteration as a smooth curve, and r3 is a random number lies in
the range of (0, 1). (2) Foraging behavior

3.2.4. Reproduction (exploitation) Underwater foraging includes fleeing and gathering behaviors.
In contrast to migration, walrus herds tend to breed in currents when
risk factors are low. During reproduction, there are mainly two behav­ a) Fleeing behavior
iors, onshore roosting, and underwater foraging. The mathematical
model is as follows. Walruses are also attacked by natural predators during underwater
foraging, and they will flee from their current activity area based on
(1) Roosting behavior danger signals from their peers. This behavior occurs in the late iteration
of the WO, and a certain degree of perturbation to the population helps
The male, female and juvenile walruses are our classification of walruses to conduct global exploration.
population members. They have different ways of renewing their ⃒ ⃒
⃒ t ⃒
position. Xi,jt+1 = Xi,jt ⋅R − ⃒Xbest − Xi,jt ⃒⋅r4 2 (17)
Step 1: Redistribution of male walruses
⃒ ⃒
The population diversity is crucial to the later iterative search for ⃒ t t ⃒
where, ⃒Xbest − Xi,j ⃒ denotes the distance between the current walrus and
superiority. In the quasi-Monte Carlo method, the Halton sequence is a
widely used method to generate randomly distributed sequences. the best walrus, r4 is a random number lies in the range of (0, 1).
Adopting Halton sequence distribution for male walrus position update
can allow a broader distribution of the population with search space. b) Gathering behavior
The principle is to divide the search area evenly into several parts and
select a random point in each part. This ensures both randomness and Walruses can cooperate to forage and move according to the location
uniformity. of other walruses in the population and sharing location information can
Step 2: Position update of female walruses help the whole walrus herd to find the sea area with higher food
The female walrus is influenced by male walrus (Maleti,j ) and the lead abundance.
t
walrus (Xbest ). As the process of iteration, the female walrus is gradually Xi,jt+1 = (X1 + X2 )/2 (18)
influenced less by the mate and more by the leader.
⎧ ⃒ ⃒
( ) ⎪ t ⃒ t ⃒
Femalet+1 t t t ⎨ X1 = Xbest − a1 × b1 × ⃒Xbest − Xi,jt ⃒
i,j = Femalei,j + α • Malei,j − Femalei,j + (1 − α)
( )
⃒ ⃒ (19)
⎪ t ⃒ t ⃒
t
• Xbest − Femaleti,j (12)
⎩ X2 = Xsecond − a2 × b2 × ⃒Xsecond − Xi,jt ⃒

where, Femalet+1 a = β × r5 − β (20)

i,j is the new position for the i th female walrus on the j th
dimension, Maleti,j and Femaleti,j are the positions of the i th male and b = tan(θ) (21)
female walruses on the j th dimension.
Step 3: Position update of juvenile walruses where, X1 and X2 are two weights affecting the gathering behavior of
Juvenile walruses at the edge of the population are often targeted by t
walrus, Xsecond is the position of the second walrus in the current itera­
⃒ ⃒
killer whales and polar bears. Therefore, juvenile walruses need to up­ ⃒ t t ⃒
tion, ⃒Xsecond − Xi,j ⃒ denotes the distance between the current walrus and
date their current position to avoid predation.
( ) the second walrus, a and b are the gathering coefficients, r5 is a random
Juvenilet+1 t
i,j = O − Juvenilei,j • P (13) number lies in the range of (0, 1), and θ takes values ranging from 0 to π.

t
O = Xbest + Juvenileti,j • LF (14) 3.3. The procedure of WO

In WO, the danger signal is used to determine whether the WO

where, Juvenilet+1
i,j is the new position for the i th juvenile walrus on the j
performs the exploration phase or the exploitation phase. When the
th dimension, Juvenileti,j is the position of the i th juvenile walrus on the j absolute value of the danger signal is not less than 1, the walrus herd
th dimension, P is the distress coefficient of juvenile walrus and is a migrates to a new domain within the solution space, which is the

6
M. Han et al. Expert Systems With Applications 239 (2024) 122413

exploration phase in the early stage of the algorithm; On the contrary, is very useful to evaluate the efficiency of an algorithm and depends on
the walrus herd reproduces, which is the exploitation phase in the late three main processes: initialization, fitness evaluation and updating of
stage of the algorithm. The security signal plays a key role in the the solution. The computational complexity of the initialization process
exploitation phase, which influences whether individual walrus choose and the updating mechanism are O(N) and O(N × T) + O(N × T × D),
roosting behavior or foraging behavior. Among them, foraging behavior where N is the population size, T is the maximum iterations, and D is the
includes two typical phenomena, gathering and fleeing, which are dimension of given problem. Therefore, computational complexity of
controlled by danger signals. The pseudo code and flowchart of WO are WO is O(N × (T +T × D +1)). The amount of memory space temporarily
described in detail in Algorithm 1 and Fig. 2, respectively. occupied by an algorithm during operation can be measured in terms of
Algorithm 1: The pseudo code of WO space complexity. The space complexity of WO algorithm is the
Input: Algorithm parameters (population size N, maximum iteration T) maximum amount of space used at any one time which is considered
1: Initialize the population and define the related parameters during its initialization process. Thus, the space complexity of WO al­
2: Evaluate the fitness values and obtain the best solution gorithm is O(N × Dim).
3: While t ≤ T
4: If |Danger signal| ≥ 1 {Exploration phase}
5: Update new position of each walrus using Eq. (9)
3.5. Walrus herd behaviors
6: Else {Exploitation phase}
7: If Safety signal ≥ 0.5 // Breeding behavior // To visualize the behavior of the proposed algorithm, the simulations
8: For each male walrus of walrus herd behavior are shown in this section. Fig. 3 depict the group
9: Update new position based on Halton sequence
behavior of a walrus herd of size 30 searching for optimal solutions of
10: End For
11: For each female walrus unimodal and multimodal functions in 3D space. The black dots indicate
12: Update new position using Eq. (12) the 30 walruses, and the red dots indicate the global optimal positions.
13: End For When t = 1, the walruses are randomly generated in the solution space
14: For each juvenile walrus with a wide scattering range. As the iteration continues, the search range
15: Update new position using Eq. (13)
16: End For
of all walruses gradually narrows. Finally, all walruses converge to the
17: Else // Foraging behavior // red dots. The search results of the unimodal function can show that WO
18: If |Danger signal| ≥ 0.5 // Gathering behavior // has good exploitation ability, and the search results of the multimodal
19: Update new position of each walrus using Eq. (17) function can reflect the strong exploration ability of WO.
20: Else // Fleeing behavior //
21: Update new position of each walrus using Eq. (18)
22: End If 4. Results and discussions
23: End If
24: End If All the algorithms are coded in MATLAB programming software and
25: Update the walrus position simulations are run on PC with Intel i7 CPU with 16 GB RAM. The codes
26: Calculate the fitness value and update the current best solution
of all comparison algorithms are published by their original authors.
27: t =t+1
28: End While Note that, all algorithms have been simulated under the same conditions
Output: the best solution (population size and maximum number of iterations) in the same class of
tests.
3.4. Computational complexity

When solving optimization problems, the computational complexity

Fig. 2. Flow chart of WO.

7
M. Han et al. Expert Systems With Applications 239 (2024) 122413

4.1. Standard benchmark functions analysis solution space, and developing the optimal location neighborhoods.
According to the trajectory in Fig. 4, we can see the sudden move­
4.1.1. Experimental setup ment of the walrus at the beginning of the search process. The amplitude
A set of benchmark functions is used to verify the optimization of these fluctuations decreases with iterations. This point ensures the
capability of the WO. This set of functions consists of three different shift of the iterative process from the exploratory trend to the devel­
groups of unimodal, multimodal, and fixed-dimension multimodal. opmental step. Finally, the movement pattern of the walrus becomes
Unimodal benchmark functions (F1-F7), which only have one best very stable. By monitoring the average fitness of all agents, we can note
global solution, are used for assessing the exploitation ability of opti­ the process of decreasing fitness values. There is a rich population di­
mization methods. Multimodal benchmark functions (F8-F23) have versity in the initial phase of the iteration. Therefore, WO can dynami­
many local optima. Thus, they are used for assessing the exploration and cally focus on more promising domains during the iterative process.
exploitation balance in metaheuristic algorithms. The features of all the Based on the convergence curves, we can observe the pattern of accel­
functions are shown in Table 2. erated decline in all curves.
Some well-known optimizers, including Artificial Bee Colony Algo­
rithm (ABC) (Karaboga & Basturk, 2007), Atomic Orbital Search (AOS) 4.1.3. Quantitative analysis
(Azizi, 2021), Butterfly Optimization Algorithm (BOA) (S. Arora & In this subsection, the performance of WO is quantified. 100 inde­
Singh, 2019), Black Widow Optimization Algorithm (BWOA) (Hay­ pendent runs of each algorithm in each test are implemented, and the
yolalam & Pourhaji Kazem, 2020), Chimp Optimization Algorithm statistical results are summarized in Table 4 and Table 5. The last three
(ChOA) (Khishe & Mosavi, 2020), Fruit fly Optimization Algorithm lines of tables are reported to display statistical analysis. The first line
(FOA) (W. T. Pan, 2012), Genetic Algorithm (GA) (Deb, 1991), Grey shows three symbols (W|T|L) that denote the number of the functions in
Wolf Optimizer (GWO) (Mirjalili et al., 2014), Moth-flame Optimization which the performance of the algorithm is the best (win) | indistin­
Algorithm (MFO) (Mirjalili, 2015a), Particle Swarm Optimization (PSO) guishable (tie) | inferior (loss) to the others. The second line illustrates
(Kennedy & Eberhart, 1995), Sand Cat Swarm Optimization (SCSO) the Friedman mean rank. The third line refers to the final rank values of
(Seyyedabbasi & Kiani, 2022), Spotted Hyena Optimizer (SHO) (Dhiman all algorithms (Abualigah et al., 2021).
& Kumar, 2017), Seagull Optimization Algorithm (SOA) (Dhiman & The unimodal functions (F1-F7) have only one global optimal solu­
Kumar, 2019), Sparrow Search Algorithm (SSA) (Xue & Shen, 2020), tion and are used to examine the ability of the algorithm to exploit the
and Whale Optimization Algorithm (WOA) (Mirjalili & Lewis, 2016) are global solution. The statistical results in Table 4 confirm the superiority
used for comparisons with the WO. The dimensions of F1-F13 are set to of the WO in the exploitation of the optimal solutions, as it obtains the
30. The maximum iterations are set to 2000 with a population size minimum of the best and worst solutions for the six functions tested in
(search agents) of 100. All the algorithms are independently run for 100 this set, as well as being smaller than the other comparison algorithms in
times. The parameters of compared algorithms are shown in Table 3. terms of average and standard deviation.
The multiple local optimal solutions of the multidimensional
4.1.2. Qualitative analysis benchmark functions (F8-F13) and the fixed-dimensional benchmark
The qualitative results of WO are evaluated using several unimodal functions (F14-F23) can test the diversification capability of the algo­
and multimodal standard criterion functions, and the results are evalu­ rithm. We can see from the results in Table 4 and Table 5 that the WO
ated using four different criteria of WO, including search history, tra­ obtains first rank both on average and standard deviation in F8-F13,
jectory of the first agent, average fitness, and convergence curve. The achieves first rank on average value in F14-F23, achieves second rank
search history graph reveals all locations visited by walrus individuals on standard deviation value in F17, and achieves third rank on standard
during the iteration; The trajectory graph monitors how the first walrus deviation in F18. BWOA achieves the best solution in F1-F4. However,
has changed during the iteration; The average fitness graph shows the the ABC, AOS, GWO, MFO, PSO, SCSO, SHO, SSA, WOA, and BWOA in
overall change in walrus population; The convergence behavior plot the comparison algorithms show a higher level in the best values in some
records the best solution after each iteration of the population. of the function tests. Compared with the WO, they fail to perform as well
Ten typical functions are selected from 23 benchmark tests for in terms of the worst, average, or standard deviation values.
analysis. It can be observed from the search history in Fig. 4 that WO The time rows in Table 4 and Table 5 record the average time ob­
tends to show a similar pattern when dealing with different situations, tained by running 100 tests independently. The statistics show that a few
including promoting diversification, exploring favorable areas of algorithms take less computation time than WO, while most of the

Fig. 3. Swarm behaviors of 30 individuals.

8
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Table 2
Summary of the unimodal benchmark functions.
Functions Description Dimensions Range fmin
∑d d
Sphere F1 = 2 [-100,100] 0
i=1 xi
∑d ∏d d
Schwefel 2.22 F2 = [-10,10] 0
i=1 |xi | + i=1 |xi |
Schwefel 1.2 ∑d (∑i )2 d [-100,100] 0
F3 = i=1 j=1 xj

Schwefel 2.21 F4 = max{|xi |, 1 ≤ i ≤ d } d [-100,100] 0

Rosenbrock ∑d− 1 [ ( )
2 2
]
d [-30,30] 0
F5 = i=1 100 xi+1 − xi +(xi − 1)2
Step ∑d d [-100,100] 0
F6 = i=1 ([xi + 0.5])2
∑ d
Quartic F7 = di=1 ix4i + rand[0, 1) [-128,128] 0
∑d √̅̅̅̅̅̅̅
Schwefel 2.26 F8 = i=1 (xi sin( |xi | )) d [-500,500] -418.9829
d
∑d [ ]
d [-5.12,5.12] 0
Rastrigin F9 = x2i − 10cos2πxi +10n
i=1
⎛ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ⎞
Ackley ( ) d [-32,32] 0
1∑d 2 ⎠ 1∑d
F10 = − 20exp⎝ − 0.2 xi − exp cos2πxi + 20 + exp(1)
d i=1 d i=1

Griewank 1 ∑d 2 ∏d xi d [-600,600] 0
F11 = x − i=1 cos √̅ + 1
4000 i=1 i i
Penalized a) π ∑d− 1 ( )2 [ ( ) ∑ ]
d [-50,50] 0
F12 = [10sin(πy1 ) ] + i=1 yi − 1 1 +10sin2 πyi+1 + di=1 u(xi , 10, 100, 4)
d ( ) ∑
∑ [ ]
F13 = 0.1 sin2 (3πx1 ) + di=1 (xi − 1)2 1 + sin2 (3πx1 + 1) +(xd − 1)2 +sin2 (2πxd ) + di=1 u(xi , 5, 100, 4) d [-50,50] 0

Foxholes 1 ∑25 1 − 1 2 [-65.536,65.536] 0.998

F14 = [ + ∑ ]
500 j=1
j + 2i=1 (xi − aij )6
Kowalik ∑11 x1 (b2i + bi x2 )
2 4 [-5,5] 0.0003075
F15 = i=1 [ai − ]
b2i + bi x3 + x4
Camel-Back 1 2 [-5,5] -1.0316
F16 = 4x21 − 2.1x41 + x61 + x1 x2 − 4x22 + 4x42
3
( )
Branin 5.1 2 5 2
1 2 [-5,5] 0.398
F17 = (x2 − x1 + x1 − 6) + 10 1 − cosx1 + 10
4π 2 π 8π
[ ]
Goldstein- F18 = 1 +(x1 + x2 + 1) (19 − 14x1 + 3x1 − 14x2 + 6x1 x2 + 3x22 ) × [30 + (2x1 − 3x2 )2 × (18 − 32xi +
2 2 2 [-2,2] 3
Price
12x21 + 48x2 − 36x1 x2 + 27x22 )]
∑4 ∑3
Hartman’s F19 = − i=1 ci exp(− i=1 aij (xj − pij )2 ) 3 [-1,2] -3.86
Family
∑4 ∑6
Shekel’s F20 = − i=1 ci exp(− i=1 aij (xj − pij )2 ) 6 [0,1] -3.32
Family F21 = −
∑5
− ai )(X − ai )T + ci ]− 1 4 [0,1] -10.1532
i=1 [(X
∑7
F22 = − i=1 [(X − ai )(X − ai )T + ci ]− 1 4 [0,1] -10.4028
∑10
F23 = − i=1 [(X − ai )(X − ai )T + ci ]− 1 4 [0,1] -10.5363
⎧
xi + 1 ⎨ K(xi − a)m if xi > a
⎪
a)
whereyi = 1 + , u(xi , a, k, m) 0 − a ≤ xi ≥ a
4 ⎩ K(− xi − a)m − axi
⎪

remaining algorithms (e.g., GWO, SOA, WOA, etc.) take more compu­ accuracy of fitness value and the fastest convergence speed.
tation time than WO. The result obtained using the Wilcoxon rank sum
test is a parameter called p-value, which measures whether the advan­ 4.1.5. Stability analysis
tage of the algorithm being measured is significant compared to the As can be seen from Fig. 6(a), most of the tested algorithms show
competing algorithms. In this test, an algorithm is statistically sub­ relatively stable performance in the unimodal benchmark functions. In­
stantial if it has a p-value smaller than 0.05. Combined with the solution dividual algorithms such as GA, MFO, ABC, SHO, WOA, and FOA exhibit
results of 100 experimental records, the “+”, “-”, and “=” symbols in discrete characteristics, with more singular values for GA and WOA,
Table 6 indicate that the optimization performance of WO is better than, indicating a large variation in the 100-test results for these algorithms.
worse than, and equal to that of the comparison algorithm, respectively. The results of multimodal benchmark functions (see Fig. 6(b)) show
In most cases, WO performs significantly better than the other com­ that the stability of most algorithms, such as ABS, AOS, and BOA, is
parison algorithms. obviously weakened, and MFO in F10 is the most obvious ones. As
shown in the test results of F14-F23 (see Fig. 6(c)), many singular values
4.1.4. Convergence analysis (red “+”) appear in BWOA, FOA, GA and SHO, indicating that their 100
The convergence curves of all algorithms in unimodal benchmark times results have the characteristics of high discreteness. In contrast,
functions for the best score obtained so far are shown in Fig. 5(a-g). All the proposed WO still maintains high stability. The optimizer can still
algorithms completed convergence in F1, F2 and F5, but WO shows the search for the global optimal value stably when a lot of local optimal
fastest convergence rate. Although FOA appears to produce the fastest value interferes. This phenomenon is consistent with the statistical re­
iteration on F1, F3, and F4-F6, its accuracy is too low compared with sults of standard deviation from Table 4 to Table 5.
other algorithms, which can be verified by combining the data in
Table 6. In the F3-F4 and F6-F7, ABC, MFO and FOA stagnate at local 4.1.6. Sensitivity analysis
extremes, which confirms that the exploitation phase of WO is efficient The process of finding the best parameter settings for a metaheuristic
and has reliable exploration value. The convergence curves of multi­ algorithm in solving a given problem is a major issue. Different values of
modal benchmark function and fixed-dimension benchmark function the control parameters of an algorithm might yield different results as
tests are shown in Fig. 5(h-m) and Fig. 5(n-w). From the curves of F8-F9, reported in the literature (Braik, 2021). Hence, there is a need to tune
F11-F13, F17 and F19-F20, the proposed WO can obtain the highest the control parameters of each metaheuristic that could provide more

9
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Table 3
Parameter values of algorithms.
Algorithms Parameters Values

ABC The percentage of onlooker bees 50%

The percentage of employed bees 50%
The number of scout bees 1
AOS Maximum number of layers around nucleus (n) 5
Photon rate for position determination of electrons (PR) 0.1
BOA Modular modality (c) 0.01
Power exponent (a) 0.1
Switch probability (p) 0.8
BWOA Procreating rate (PP) 0.6
Cannibalism rate (CR) 0.44
Mutation rate (Pm) 0.4
ChOA m chaotic vector
GA Crossover percentage (pc) 0.8
Mutation rate (pm) 0.1
MFO A constant of the logarithmic spiral (b) 1
PSO The maximum and minimum of inertia weight (ωmax , ωmin ) 0.9, 0.6
c1 , c 2 2
SCSO Simulated hearing characteristics of the sand cat (SM ) 2
SOA Control parameters (fc ) 2
Constants to define the spiral shape (u, v) 1
SSA Initial speed (v0 ) 0
WOA A constant of the logarithmic spiral (b) 1
AFT Initial estimate of the tracking distance at the first iteration (α0 ) 1
Final rough estimation of the probability at the end iteration (β0 ) 0.1
Constant values that control exploration and exploitation ability (α1 , β1 ) 2
ALO Constant defined based on the current iteration (w) 2, 3, 4, 5
CSA Probability of a chameleon finding its prey (Pp ) 0.1
Three constant values that control exploration and exploitation ability (γ, α, β) 1, 3.5, 3
Two positive constants that control the velocity of the chameleon tongue (c1 , c2 ) 1.75
Two positive numbers that control exploration ability (p1 , p2 ) 0.25, 1.75
A positive number that controls exploitation ability (ρ) 1
DMOA The alpha female’s vocalization that keeps the family within a path (peep) 2
Number of babysitters 3
GOA The intensity and length scale of attraction (f, l) 0.5, 1.5
The upper and lower limits of the decreasing coefficient (cmax , cmin ) 1, 0.00004
MVO Wormhole existence probability (WEPmax , WEPmin ) 1, 0.2
The exploitation accuracy over the iterations (p) 6
NGO A random number I 1 or 2
SCA A constant for movement direction (a) 2
Three constants for food quantity, exploration, and exploitation (c1 , c2 , c3 ) 0.5, 0.05, 2
Threshold for food quantity 0.25, 0.6
SS Initial speed (v0 ) 0
WSO The acceleration coefficient (τ) 4.125
The upper and lower frequencies of the undulating motion (fmax , fmin ) 0.75, 0.07
Three positive constants employed to manage exploration and exploitation behaviors (a0 , a1 , a2 ) 6.25, 100, 0.0005

adaptability and robustness in solving a broad range of problems in iterations increases. Moreover, it can be seen from Fig. 7(b) that the
various fields of science. The optimal parameter set of the proposed number of iterations decreases when the number of search agents in­
algorithm was selected based on the use of Design of Experiment (DoE) creases. Fig. 7(c) displays relatively stable behavior of WO when p
framework through empirical testing of the proposed algorithm on a ranges from 0.3 to 0.4. Based on this analysis, the best value for p, with
selected set of test functions. WO having the best results, can be suggested as its largest value (i.e., p =
Here, the proposed parameter setting method examines the sensi­ 0.45). In F14 and F15 tests, WO has relatively slight sensitivity to these
tivity of the maximum number of iterations (T), number of search agents parameters, where it presented small differences between the results.
(N) and the proportion of male walruses to the walrus herd (p). F5, F6,
F12, F13, F14 and F15 from the unimodal and multimodal benchmark 4.1.7. Scalability analysis
functions were selected for testing. The values of each parameter for the The scalability of the proposed WO is tested by varying the di­
sensitivity analysis design were defined as follows: T = {200,500,1000, mensions of F1-F13 to determine the effect of increasing the dimensions
2000}, N = { 30, 50, 80, 100}, p = {0.3, 0.35, 0.4, 0.45}. Each function on the computational results of different test functions. The dimensions
was solved with 30 independent runs. The results of the sensitivity of the test functions are 50, 100, and 500. The 15 algorithms in Section
analysis, in terms of the mean fitness and convergence curves, for the 4.1.1 (function dimension is 30) are selected for comparative analysis.
above four functions with the three control parameters are illustrated This reveals the effect of dimensionality change on the convergence
below. ability of WO, which is used to diagnose the effectiveness of WO for low
Table 10 shows the average fitness of WO when used to simulate six and high dimensional problems.
different benchmark functions. Fig. 7(a) and (b) show the WO conver­ The scalability analysis included 39 tests (13 standard functions).
gence curves of F1 under different maximum iterations and number of From Tables 7 to 9, the WO outperforms the comparison algorithms in
search agents, respectively. Fig. 7(c) displays the convergence curve of most cases, with 49 out of 52 cases (94.230 %) achieving the best mean
WO on F5 with the different proportion of male walruses to the walrus and standard deviation values, which is higher than BWOA (53.846 %),
herd. The computational results and the visualization demonstrate that SCSO (38.462 %), WOA (17.308 %), SHO (13.462 %), SOA (5.769 %),
WO converges to the optimum solution when the maximum number of AOS (3.846 %), BOA (1.923 %), and other algorithms (0 %). In addition,

10
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Fig. 4. Qualitative results of WO.

11
 M. Han et al.
Table 4
Results of unimodal and multimodal benchmark functions (F1-F13) with d = 30.
Comparative algorithms

Function ABC AOS BOA BWOA ChOA FOA GA GWO MFO PSO SCSO SHO SOA SSA WOA WO

F1 Best 2.42E-10 1.74E-266 1.64E-17 0.00Eþ00 2.00E-109 6.80E-04 8.35E-03 1.33E-178 6.73E-17 5.69E-05 0.00Eþ00 0.00Eþ00 1.13E-57 3.26E-09 0.00Eþ00 0.00Eþ00
Worst 1.33E-08 1.12E-248 2.13E-17 0.00Eþ00 4.23E-57 1.19E-03 3.31E-02 8.36E-173 1.00E+04 4.36E-03 0.00Eþ00 1.93E-278 3.56E-48 7.06E-09 0.00Eþ00 0.00Eþ00
Avg 2.12E-09 1.12E-250 1.93E-17 0.00Eþ00 5.87E-59 9.18E-04 1.61E-02 3.64E-174 1.80E+03 7.55E-04 0.00Eþ00 1.93E-280 5.00E-50 5.09E-09 0.00Eþ00 0.00Eþ00
Std 1.80E-09 0.00E+00 9.38E-19 0.00Eþ00 4.47E-58 1.17E-04 4.65E-03 0.00E+00 3.86E+03 7.06E-04 0.00Eþ00 0.00Eþ00 3.58E-49 8.64E-10 0.00Eþ00 0.00Eþ00
Time 4.06E+00 1.91E+00 5.62E-01 2.32E+00 2.52E+01 4.59E-01 4.81E-01 1.20E+00 7.03E-01 4.29E-01 1.13E+01 2.06E+00 2.08E+00 1.08E+00 1.78E+01 8.33E-01
(s)
F2 Best 1.76E-08 6.75E-138 3.58E-15 0.00Eþ00 9.16E-65 1.36E+00 1.80E-01 1.22E-101 3.35E-11 2.84E-03 3.81E-274 0.00Eþ00 5.77E-42 3.30E-05 6.75E-243 0.00Eþ00
Worst 2.79E-06 5.98E-129 1.67E-14 0.00Eþ00 4.25E-37 1.90E+00 4.47E-01 2.44E-98 1.10E+02 5.47E-02 5.07E-263 6.31E-146 2.38E-37 2.69E+00 6.71E-222 0.00Eþ00
Avg 2.63E-07 1.31E-130 1.51E-14 0.00Eþ00 5.81E-39 1.59E+00 3.19E-01 1.22E-99 2.96E+01 1.43E-02 7.15E-265 6.31E-148 1.11E-38 2.08E-01 7.61E-224 0.00Eþ00
Std 4.37E-07 7.43E-130 1.74E-15 0.00Eþ00 4.31E-38 1.15E-01 4.70E-02 2.84E-99 2.19E+01 7.87E-03 0.00Eþ00 6.31E-147 3.13E-38 4.54E-01 0.00Eþ00 0.00Eþ00
Time 4.61E+00 2.14E+00 6.11E-01 2.97E+00 2.39E+01 5.65E-01 5.03E-01 1.20E+00 7.29E-01 4.32E-01 1.13E+01 1.95E+00 2.18E+00 1.11E+00 1.70E+01 8.31E-01
(s)
F3 Best 2.39E+04 3.74E-216 1.72E-17 0.00Eþ00 4.64E-31 1.86E-01 4.19E-01 2.65E-66 2.20E-01 9.26E+00 0.00Eþ00 2.80E-02 6.50E-14 8.23E-08 1.90E+00 0.00Eþ00
Worst 6.12E+04 1.87E-193 2.24E-17 0.00Eþ00 1.89E-16 3.80E-01 3.26E+03 1.15E-53 4.50E+04 3.22E+01 0.00Eþ00 2.11E+04 1.75E-05 7.33E-06 5.89E+03 0.00Eþ00
Avg 4.34E+04 1.89E-195 2.02E-17 0.00Eþ00 2.19E-18 2.57E-01 2.44E+02 1.97E-55 1.36E+04 1.80E+01 0.00Eþ00 4.10E+03 7.77E-07 4.39E-07 6.25E+02 0.00Eþ00
Std 7.75E+03 0.00Eþ00 9.95E-19 0.00Eþ00 1.89E-17 3.85E-02 4.89E+02 1.35E-54 1.20E+04 4.89E+00 0.00Eþ00 4.94E+03 2.70E-06 8.17E-07 8.04E+02 0.00Eþ00
Time 6.98E+00 3.53E+00 3.17E+00 9.05E+00 2.52E+01 1.92E+00 1.75E+00 2.62E+00 1.97E+00 1.66E+00 1.26E+01 4.58E+00 3.71E+00 2.43E+00 1.67E+01 2.36E+00
(s)
F4 Best 3.21E+01 3.06E-129 1.43E-14 0.00Eþ00 2.64E-23 6.20E-03 1.01E-01 9.47E-47 2.05E+00 4.88E-01 8.15E-226 0.00Eþ00 1.57E-12 3.50E-05 5.08E-13 0.00Eþ00
Worst 4.80E+01 7.24E-121 1.83E-14 0.00Eþ00 1.80E-12 1.01E-02 2.07E-01 3.94E-42 4.08E+01 1.26E+00 6.14E-212 8.52E-73 1.87E-07 3.70E+00 8.13E+01 0.00Eþ00
Avg 4.11E+01 1.94E-122 1.64E-14 0.00Eþ00 2.77E-14 8.20E-03 1.57E-01 8.73E-44 1.42E+01 8.77E-01 7.88E-214 8.52E-75 4.00E-09 1.25E-01 1.33E+01 0.00Eþ00
Std 3.28E+00 7.87E-122 7.52E-16 0.00Eþ00 1.82E-13 7.85E-04 1.89E-02 4.17E-43 8.62E+00 1.48E-01 0.00Eþ00 8.52E-74 1.97E-08 4.22E-01 2.05E+01 0.00Eþ00
Time 4.12E+00 2.09E+00 5.60E-01 2.66E+00 2.55E+01 4.93E-01 4.73E-01 1.28E+00 6.92E-01 4.19E-01 1.13E+01 2.08E+00 2.61E+00 1.22E+00 1.53E+01 8.75E-01
(s)
12

F5 Best 4.98E+01 2.43E+01 2.88E+01 2.88E+01 2.65E+01 2.82E+01 9.50E+00 2.41E+01 1.80E+00 2.23E+01 2.50E+01 2.69E-04 2.64E+01 9.30E+00 3.05E-02 2.66E-08
Worst 1.18E+03 2.53E+01 2.90E+01 2.90E+01 2.90E+01 2.91E+01 8.20E+01 2.85E+01 9.01E+04 5.29E+02 2.88E+01 2.80E+01 2.87E+01 2.77E+02 2.60E+01 2.78E-03
Avg 2.56E+02 2.48E+01 2.89E+01 2.89E+01 2.87E+01 2.88E+01 3.38E+01 2.59E+01 2.11E+04 9.60E+01 2.68E+01 1.30E+01 2.73E+01 4.30E+01 2.48E+01 2.91E-04
Std 1.70E+02 1.94E-01 2.60E-02 3.96E-02 4.13E-01 1.85E-01 2.11E+01 8.57E-01 3.79E+04 7.34E+01 1.11E+00 1.37E+01 5.32E-01 3.69E+01 2.51E+00 5.07E-04
Time 4.62E+00 2.20E+00 8.40E-01 2.92E+00 2.59E+01 6.89E-01 6.12E-01 1.44E+00 8.84E-01 5.41E-01 1.14E+01 2.22E+00 2.53E+00 1.42E+00 1.60E+01 9.46E-01
(s)
F6 Best 4.16E-10 1.22E-04 3.56E+00 2.26E+00 1.22E+00 7.64E+00 7.37E+00 8.70E-07 5.00E-17 4.72E-05 1.60E-07 4.53E-05 1.04E-02 2.46E-09 8.05E-06 4.39E-11
Worst 2.93E-08 4.61E-04 5.94E+00 6.83E+00 2.60E+00 7.69E+00 8.05E+00 1.01E+00 1.98E+04 3.93E-03 1.01E+00 1.01E+00 7.56E-01 7.90E-09 5.76E-05 1.58E-07
Avg 2.29E-09 2.71E-04 4.86E+00 4.71E+00 1.88E+00 7.66E+00 7.85E+00 1.29E-01 1.10E+03 6.80E-04 2.62E-01 1.12E-01 2.28E-01 5.24E-09 2.52E-05 3.59E-08
Std 3.18E-09 6.46E-05 5.93E-01 9.62E-01 2.87E-01 1.07E-02 1.01E-01 1.91E-01 3.43E+03 7.36E-04 2.18E-01 1.87E-01 1.72E-01 9.31E-10 1.02E-05 3.49E-08
Time 3.90E+00 1.99E+00 5.22E-01 2.51E+00 2.48E+01 4.80E-01 4.52E-01 1.22E+00 7.14E-01 4.12E-01 1.10E+01 2.04E+00 2.46E+00 1.34E+00 1.72E+01 8.08E-01
(s)
F7 Best 1.92E-02 3.05E-05 7.18E-05 9.84E-07 8.27E-06 7.85E+00 3.81E-01 1.68E-05 4.91E-03 6.72E-02 5.83E-08 7.90E-07 1.85E-03 5.90E-03 1.43E-06 2.86E-07
Worst 7.78E-02 3.90E-04 4.07E-04 5.54E-05 8.61E-04 4.95E+01 1.51E+00 4.13E-04 1.88E+01 4.94E-01 1.19E-04 3.16E-04 2.05E-02 4.84E-02 1.02E-03 7.42E-05

Expert Systems With Applications 239 (2024) 122413

Avg 4.65E-02 1.42E-04 2.26E-04 1.19E-05 1.47E-04 2.14E+01 8.56E-01 1.23E-04 1.68E+00 1.93E-01 1.18E-05 2.22E-05 8.44E-03 1.59E-02 2.03E-04 1.80E-05
Std 1.13E-02 7.50E-05 8.01E-05 9.71E-06 1.63E-04 7.50E+00 2.29E-01 7.87E-05 4.30E+00 8.26E-02 1.54E-05 3.43E-05 3.71E-03 6.71E-03 2.28E-04 1.83E-05
Time 5.48E+00 2.87E+00 2.05E+00 3.16E+00 2.75E+01 1.30E+00 1.20E+00 1.96E+00 1.49E+00 1.15E+00 1.18E+01 3.60E+00 3.13E+00 2.10E+00 1.94E+01 1.65E+00
(s)
F8 Best -3.49E+243 -9.15E+03 -4.10E+03 -7.80E+03 -6.00E+03 -3.95E+00 -3.92E+03 -8.51E+03 -1.09E+04 -9.87E+03 -9.14E+03 -1.26Eþ04 -6.07E+03 -9.31E+03 -1.26Eþ04 -1.26Eþ04
Worst -1.58E+233 -6.24E+03 -6.33E+03 -3.95E+03 -5.75E+03 -7.74E-01 -1.59E+03 -4.10E+03 -7.22E+03 -5.09E+03 -5.05E+03 -1.26Eþ04 -4.81E+03 -6.08E+03 -8.56E+03 -1.26Eþ04
Avg -5.07E+241 -7.78E+03 -4.78E+03 -5.75E+03 -5.84E+03 -2.48E+00 -2.66E+03 -6.45E+03 -9.00E+03 -6.81E+03 -7.27E+03 -1.26Eþ04 -5.31E+03 -7.73E+03 -1.22E+04 -1.26Eþ04
Std 6.55E+04 6.58E+02 3.65E+02 7.90E+02 5.93E+01 9.57E-01 5.04E+02 6.60E+02 8.50E+02 8.56E+02 7.83E+02 3.24E-01 2.74E+02 7.30E+02 7.26E+02 5.44E-04
Time 6.12E+00 2.25E+00 1.73E+00 2.74E+00 2.63E+01 6.26E-01 1.08E-02 1.37E+00 8.72E-01 6.12E-01 1.19E+01 2.43E+00 2.59E+00 1.40E+00 1.79E+01 1.02E+00
(s)
F9 Best 1.56E+02 0.00Eþ00 0.00Eþ00 0.00Eþ00 0.00Eþ00 4.59E+01 7.66E-01 0.00Eþ00 3.58E+01 2.31E+01 0.00Eþ00 0.00Eþ00 0.00Eþ00 1.09E+01 0.00Eþ00 0.00Eþ00
Worst 2.17E+02 6.62E+01 1.89E+02 0.00Eþ00 3.45E+00 8.71E+01 3.15E+00 0.00Eþ00 2.38E+02 9.93E+01 0.00Eþ00 0.00Eþ00 0.00Eþ00 8.86E+01 0.00Eþ00 0.00Eþ00
Avg 1.95E+02 1.19E+01 1.89E+00 0.00Eþ00 4.40E-02 6.52E+01 1.82E+00 0.00Eþ00 1.18E+02 4.86E+01 0.00Eþ00 0.00Eþ00 0.00Eþ00 4.26E+01 0.00Eþ00 0.00Eþ00
Std 1.12E+01 1.71E+01 1.89E+01 0.00Eþ00 3.57E-01 8.41E+00 5.46E-01 0.00Eþ00 3.82E+01 1.35E+01 0.00Eþ00 0.00Eþ00 0.00Eþ00 1.57E+01 0.00Eþ00 0.00Eþ00
(continued on next page)
 M. Han et al.
Table 4 (continued )
Comparative algorithms

Function ABC AOS BOA BWOA ChOA FOA GA GWO MFO PSO SCSO SHO SOA SSA WOA WO

Time 4.16E+00 2.21E+00 8.30E-01 2.46E+00 2.46E+01 5.84E-01 5.28E-01 1.21E+00 8.14E-01 5.50E-01 1.13E+01 1.99E+00 2.49E+00 1.35E+00 1.77E+01 9.16E-01
(s)
F10 Best 1.17E-04 8.88E-16 4.44E-15 8.88E-16 2.00E+01 6.64E-02 5.19E-02 4.44E-15 2.66E-09 4.42E-03 8.88E-16 8.88E-16 4.44E-15 1.22E-05 8.88E-16 8.88E-16
Worst 9.98E-03 4.44E-15 1.51E-14 8.88E-16 2.00E+01 1.06E-01 1.21E-01 1.51E-14 2.00E+01 8.66E-02 8.88E-16 7.99E-15 7.99E-15 2.89E+00 7.99E-15 8.88E-16
Avg 1.61E-03 4.12E-15 1.04E-14 8.88E-16 2.00E+01 8.55E-02 8.41E-02 8.03E-15 9.65E+00 1.81E-02 8.88E-16 1.21E-15 4.48E-15 1.30E+00 3.62E-15 8.88E-16
Std 1.42E-03 1.02E-15 3.60E-15 0.00Eþ00 1.54E-03 7.56E-03 1.60E-02 9.44E-16 9.52E+00 1.10E-02 0.00Eþ00 1.14E-15 3.55E-16 9.00E-01 2.47E-15 0.00Eþ00
Time 4.46E+00 2.22E+00 6.59E-01 2.47E+00 2.41E+01 5.70E-01 5.54E-01 1.19E+00 8.18E-01 5.47E-01 1.10E+01 2.18E+00 2.45E+00 1.35E+00 1.85E+01 9.42E-01
(s)
F11 Best 8.53E-08 0.00Eþ00 0.00Eþ00 0.00Eþ00 0.00Eþ00 9.66E-07 2.71E-04 0.00Eþ00 1.11E-16 5.07E-06 0.00Eþ00 0.00Eþ00 0.00Eþ00 8.98E-09 0.00Eþ00 0.00Eþ00
Worst 1.61E-02 9.92E-03 0.00Eþ00 0.00Eþ00 1.05E-02 2.39E-06 1.24E-03 1.52E-02 9.07E+01 5.67E-02 0.00Eþ00 7.82E-01 0.00Eþ00 3.44E-02 3.17E-02 0.00Eþ00
Avg 3.32E-04 5.21E-04 0.00Eþ00 0.00Eþ00 2.84E-04 1.48E-06 5.91E-04 4.52E-04 4.53E+00 1.41E-02 0.00Eþ00 7.82E-03 0.00Eþ00 8.89E-03 9.95E-04 0.00Eþ00
Std 1.68E-03 2.09E-03 0.00Eþ00 0.00Eþ00 1.64E-03 2.90E-07 1.75E-04 2.30E-03 1.98E+01 1.21E-02 0.00Eþ00 7.82E-02 0.00Eþ00 9.46E-03 4.57E-03 0.00Eþ00
Time 4.65E+00 2.34E+00 9.84E-01 2.76E+00 2.36E+01 7.31E-01 6.58E-01 1.26E+00 9.36E-01 6.09E-01 1.14E+01 2.44E+00 2.54E+00 1.46E+00 1.74E+01 1.05E+00
(s)
13

F12 Best 1.68E+00 1.71E-05 1.10E-01 1.30E-01 7.90E-02 1.71E+00 1.45E+00 7.40E-08 5.15E-17 2.38E-07 2.70E-08 2.88E-07 9.41E-04 1.05E-11 1.43E-06 1.27E-12
Worst 1.47E+01 5.68E-05 4.86E-01 9.00E-01 6.72E-01 1.73E+00 1.73E+00 3.95E-02 1.77E+00 2.22E-05 5.31E-02 4.75E-02 5.44E-02 7.90E+00 1.20E-05 2.58E-09
Avg 6.87E+00 3.53E-05 2.86E-01 4.35E-01 1.73E-01 1.72E+00 1.69E+00 1.28E-02 1.82E-01 4.08E-06 2.12E-02 1.03E-02 1.21E-02 1.89E+00 4.22E-06 3.42E-10
Std 2.50E+00 9.14E-06 8.29E-02 1.45E-01 1.15E-01 3.66E-03 4.39E-02 9.31E-03 3.48E-01 4.43E-06 1.14E-02 1.05E-02 8.00E-03 1.75E+00 1.96E-06 4.72E-10
Time 8.50E+00 4.16E+00 5.07E+00 6.17E+00 2.54E+01 2.70E+00 2.65E+00 3.24E+00 2.84E+00 2.56E+00 1.30E+01 6.18E+00 4.76E+00 3.51E+00 2.21E+01 3.10E+00
(s)
F13 Best 1.25E+00 2.57E-04 7.90E-01 1.47E+00 2.27E+00 2.79E+00 8.36E-04 1.38E-06 1.98E-16 5.87E-06 3.79E-07 3.04E-07 2.09E+00 1.32E-10 1.73E-05 2.37E-12
Worst 1.87E+01 2.16E-02 2.91E+00 3.00E+00 3.00E+00 2.87E+00 4.40E-03 5.00E-01 3.60E+00 1.13E-02 2.60E+00 8.04E-01 2.48E+00 2.10E-02 1.11E-02 1.90E-07
Avg 6.72E+00 4.67E-03 1.73E+00 2.81E+00 2.85E+00 2.83E+00 1.93E-03 1.55E-01 1.56E-01 9.67E-04 1.54E+00 1.80E-01 2.33E+00 4.87E-03 1.20E-03 1.52E-08
Std 4.18E+00 6.33E-03 4.44E-01 3.63E-01 1.45E-01 1.89E-02 7.13E-04 1.24E-01 6.39E-01 2.62E-03 6.75E-01 2.25E-01 7.50E-02 6.87E-03 3.30E-03 2.56E-08
Time 8.42E+00 4.28E+00 4.92E+00 6.04E+00 2.55E+01 2.65E+00 2.55E+00 3.17E+00 2.85E+00 2.58E+00 1.27E+01 6.24E+00 4.09E+00 3.47E+00 2.25E+01 3.07E+00
(s)
(W|T|L) (0|9|4) (3|10|0) (2|11|0) (3|10|0) (2|10|1) (0|7|6) (0|11|2) (2|11|0) (3|10|0) (0|13|0) (4|9|0) (3|10|0) (2|11|0) (0|13|0) (3|10|0) (9|4|0)

Expert Systems With Applications 239 (2024) 122413

Friedman mean 12.15 5.81 9.35 5.69 9.08 12.38 11.92 6.62 13.69 10.92 4.42 5.81 8.69 11.46 6.00 2.00
rank
Final rank 14 4 10 3 9 15 13 7 16 11 2 4 8 12 6 1
 M. Han et al.
Table 5
Results of fixed-dimension multimodal benchmark functions (F14-F23).
Comparative algorithms

Function ABC AOS BOA BWOA ChOA FOA GA GWO MFO PSO SCSO SHO SOA SSA WOA WO

F14 Best 9.98E-01 9.98E-01 9.98E-01 9.98E-01 9.98E-01 1.27E+01 9.99E-01 9.98E-01 9.98E-01 9.98E-01 9.98E-01 9.98E-01 9.98E-01 9.98E-01 9.98E-01 9.98E-01
Worst 9.98E-01 1.99E+00 9.98E-01 4.95E+00 2.98E+00 1.27E+01 1.27E+01 1.08E+01 9.98E-01 1.99E+00 2.98E+00 2.98E+00 1.01E+00 9.98E-01 2.98E+00 9.98E-01
Avg 9.98E-01 1.01E+00 9.98E-01 1.45E+00 1.02E+00 1.27E+01 8.89E+00 2.61E+00 9.98E-01 1.17E+00 1.51E+00 1.20E+00 9.98E-01 9.98E-01 1.02E+00 9.98E-01
Std 3.52E-11 9.94E-02 1.26E-05 1.02E+00 1.98E-01 2.83E-11 3.98E+00 3.01E+00 2.57E-15 3.75E-01 8.75E-01 5.98E-01 6.63E-04 2.03E-15 1.98E-01 1.80E-15
Time (s) 1.34E+01 6.03E+00 8.16E+00 1.07E+01 5.51E+00 4.56E+00 4.33E+00 4.21E+00 4.37E+00 4.07E+00 4.80E+00 8.47E+00 3.98E+00 4.52E+00 8.54E+00 4.56E+00
F15 Best 7.77E-04 3.08E-04 3.08E-04 3.08E-04 1.22E-03 3.08E-04 3.08E-04 3.08E-04 3.08E-04 3.08E-04 3.08E-04 3.10E-04 3.08E-04 3.08E-04 3.08E-04 3.08E-04
Worst 1.17E-03 1.22E-03 3.42E-04 2.04E-02 1.29E-03 1.21E-03 9.02E-03 2.04E-02 1.49E-03 1.06E-03 1.22E-03 2.59E-03 1.22E-03 1.33E-03 1.23E-03 3.08E-04
Avg 1.04E-03 4.74E-04 3.14E-04 2.40E-03 1.24E-03 3.38E-04 6.79E-04 1.76E-03 7.38E-04 5.80E-04 3.55E-04 9.60E-04 9.28E-04 7.31E-04 4.83E-04 3.08E-04
Std 5.07E-05 3.53E-04 5.09E-06 6.03E-03 1.28E-05 1.00E-04 1.11E-03 5.13E-03 3.63E-04 2.93E-04 2.01E-04 7.60E-04 4.07E-04 3.65E-04 3.05E-04 4.57E-08
Time (s) 4.43E+00 1.72E+00 5.28E-01 2.63E+00 3.86E+00 3.95E-01 3.99E-01 3.11E-01 2.78E-01 1.27E-01 1.57E+00 4.30E-01 4.18E-01 4.36E-01 4.20E+00 3.94E-01
F16 Best -9.53E-01 -1.03Eþ00 -9.54E-01 -1.03Eþ00 -1.03Eþ00 -9.56E-01 -1.03E+00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00
Worst -6.49E-01 -1.03Eþ00 -6.91E-01 -1.03Eþ00 -1.03Eþ00 -6.42E-01 -3.51E-02 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00
Avg -8.92E-01 -1.03Eþ00 -8.94E-01 -1.03Eþ00 -1.03Eþ00 -8.99E-01 -6.62E-01 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00 -1.03Eþ00
Std 5.68E-02 3.35E-09 6.02E-02 3.71E-05 1.02E-06 5.91E-02 2.93E-01 4.44E-10 1.56E-15 1.56E-15 3.00E-12 8.93E-10 2.45E-09 2.43E-15 3.01E-14 1.46E-15
Time (s) 4.64E+00 1.73E+00 7.74E-01 2.37E+00 1.76E+00 3.87E-01 1.06E-02 2.64E-01 2.59E-01 1.24E-01 9.05E-01 3.36E-01 3.22E-01 4.15E-01 3.11E+00 3.70E-01
F17 Best 2.83E+00 3.98E-01 3.90E+00 3.98E-01 3.98E-01 4.24E+00 4.03E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01
Worst 4.24E+01 3.98E-01 4.41E+01 3.98E-01 3.99E-01 4.15E+01 2.68E+00 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01
Avg 1.64E+01 3.98E-01 1.59E+01 3.98E-01 3.98E-01 1.51E+01 8.39E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01 3.98E-01
Std 1.25E+01 5.95E-06 1.26E+01 1.06E-15 1.45E-04 1.20E+01 4.20E-01 1.07E-05 1.06E-15 1.06E-15 2.64E-10 1.66E-05 1.33E-07 1.60E-15 1.34E-09 1.14E-15
Time (s) 4.04E+00 1.83E+00 4.43E-01 2.27E+00 1.66E+00 3.61E-01 3.30E-01 2.00E-01 2.01E-01 6.41E-02 8.37E-01 2.26E-01 2.58E-01 4.18E-01 3.25E+00 3.25E-01
F18 Best 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00 8.40E+01 3.01E+00 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00
Worst 3.00Eþ00 3.00Eþ00 3.04E+00 3.00E+01 3.00Eþ00 8.04E+02 8.21E+01 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.49E+01 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00
Avg 3.00Eþ00 3.00Eþ00 3.00E+00 3.27E+00 3.00Eþ00 1.10E+02 1.70E+01 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00 6.12E+00 3.00Eþ00 3.00Eþ00 3.00Eþ00 3.00Eþ00
Std 9.48E-16 5.33E-08 4.47E-03 2.70E+00 1.28E-06 1.18E+02 1.58E+01 2.80E-07 4.25E-15 8.43E-16 4.98E-08 8.94E+00 5.95E-08 2.46E-14 2.36E-08 1.41E-15
Time (s) 3.97E+00 1.69E+00 4.17E-01 2.37E+00 1.65E+00 3.31E-01 3.23E-01 1.92E-01 1.94E-01 6.34E-02 8.16E-01 2.10E-01 2.57E-01 3.40E-01 3.36E+00 3.01E-01
14

F19 Best -3.86Eþ00 -3.86Eþ00 -3.82E+00 -3.86Eþ00 -3.86Eþ00 -3.82E+00 -3.85E+00 -3.86Eþ00 -3.86Eþ00 -3.86Eþ00 -3.86Eþ00 -3.86Eþ00 -3.86E+00 -3.86Eþ00 -3.86Eþ00 -3.86Eþ00
Worst -3.86Eþ00 -3.86E+00 -2.48E+00 -3.86E+00 -3.85E+00 -2.61E+00 -2.85E+00 -3.86E+00 -3.86Eþ00 -3.86Eþ00 -3.86E+00 -3.09E+00 -3.86E+00 -3.86Eþ00 -3.86E+00 -3.86Eþ00
Avg -3.86Eþ00 -3.86Eþ00 -3.71E+00 -3.86Eþ00 -3.86E+00 -3.70E+00 -3.59E+00 -3.86E+00 -3.86Eþ00 -3.86Eþ00 -3.86E+00 -3.84E+00 -3.86E+00 -3.86Eþ00 -3.86E+00 -3.86Eþ00
Std 6.25E-15 7.88E-04 2.08E-01 7.88E-04 1.71E-03 1.93E-01 2.11E-01 2.11E-03 6.25E-15 6.25E-15 2.66E-03 1.11E-01 2.38E-04 6.71E-15 1.44E-03 3.28E-15
Time (s) 4.28E+00 1.79E+00 1.69E+00 3.39E+00 2.50E+00 4.32E-01 1.08E-02 3.32E-01 3.12E-01 1.68E-01 1.27E+00 4.47E-01 4.06E-01 4.76E-01 4.38E+00 4.35E-01
F20 Best -3.32E+00 -3.32E+00 -1.63E+00 -3.32E+00 -3.31E+00 -1.84E+00 -3.05E+00 -3.32E+00 -3.32E+00 -3.32E+00 -3.32E+00 -3.31E+00 -3.09E+00 -3.32E+00 -3.32E+00 -3.32Eþ00
Worst -3.20E+00 -3.13E+00 -3.66E-01 -3.04E+00 -2.27E+00 -4.95E-01 -1.07E+00 -3.09E+00 -3.20E+00 -3.20E+00 -2.27E+00 -2.37E+00 -1.43E+00 -3.20E+00 -3.09E+00 -3.32Eþ00
Avg -3.27E+00 -3.26E+00 -9.51E-01 -3.26E+00 -2.87E+00 -9.58E-01 -2.05E+00 -3.26E+00 -3.23E+00 -3.26E+00 -3.24E+00 -3.13E+00 -2.78E+00 -3.22E+00 -3.26E+00 -3.32Eþ00
Std 5.96E-02 6.98E-02 2.97E-01 6.98E-02 2.73E-01 2.69E-01 4.36E-01 6.64E-02 4.87E-02 5.97E-02 1.21E-01 1.36E-01 3.95E-01 4.02E-02 6.65E-02 2.21E-15
Time (s) 4.33E+00 1.83E+00 1.73E+00 3.34E+00 4.71E+00 4.51E-01 1.07E-02 4.44E-01 3.89E-01 2.08E-01 2.34E+00 6.28E-01 5.85E-01 5.50E-01 5.86E+00 4.85E-01
F21 Best -1.02Eþ01 -1.02Eþ01 -9.47E+00 -1.02Eþ01 -5.05E+00 -8.56E-01 -3.01E+00 -1.02Eþ01 -1.02Eþ01 -1.02Eþ01 -1.02Eþ01 -1.02Eþ01 -5.06E+00 -1.02Eþ01 -1.02Eþ01 -1.02Eþ01
Worst -1.02Eþ01 -5.06E+00 -4.87E+00 -2.63E+00 -8.81E-01 -4.58E-01 -3.39E-01 -5.06E+00 -2.63E+00 -5.06E+00 -5.06E+00 -8.80E-01 -6.19E-01 -5.06E+00 -1.02Eþ01 -1.02Eþ01
Avg -1.02Eþ01 -9.80E+00 -6.54E+00 -9.19E+00 -4.57E+00 -5.44E-01 -8.44E-01 -9.70E+00 -8.18E+00 -8.73E+00 -6.84E+00 -8.56E+00 -2.39E+00 -9.09E+00 -1.02Eþ01 -1.02Eþ01

Expert Systems With Applications 239 (2024) 122413

Std 2.07E-14 1.29E+00 1.49E+00 2.10E+00 1.30E+00 7.75E-02 4.83E-01 1.46E+00 2.76E+00 2.29E+00 2.44E+00 2.69E+00 1.54E+00 2.07E+00 5.51E-05 1.79E-14
Time (s) 4.39E+00 1.89E+00 1.45E+00 3.35E+00 3.34E+00 5.08E-01 1.08E-02 4.37E-01 4.11E-01 2.30E-01 1.69E+00 6.10E-01 5.22E-01 6.01E-01 4.69E+00 5.46E-01
F22 Best -1.04E+01 -1.04E+01 -9.68E+00 -1.04E+01 -8.90E+00 -8.08E-01 -4.27E+00 -1.04E+01 -1.04E+01 -1.04E+01 -1.04E+01 -1.04E+01 -5.46E+00 -1.04E+01 -1.04E+01 -1.04Eþ01
Worst -1.04E+01 -5.09E+00 -4.79E+00 -1.84E+00 -9.08E-01 -4.66E-01 -5.14E-01 -5.09E+00 -2.75E+00 -5.09E+00 -5.09E+00 -2.28E+00 -7.27E-01 -2.77E+00 -2.77E+00 -1.04Eþ01
Avg -1.04E+01 -1.02E+01 -6.64E+00 -8.45E+00 -5.09E+00 -5.57E-01 -9.37E-01 -1.02E+01 -9.20E+00 -9.56E+00 -7.00E+00 -8.83E+00 -2.35E+00 -1.00E+01 -1.03E+01 -1.04Eþ01
Std 1.60E-14 1.05E+00 1.46E+00 2.78E+00 6.66E-01 5.04E-02 4.51E-01 9.11E-01 2.54E+00 1.95E+00 2.56E+00 2.44E+00 1.53E+00 1.46E+00 9.26E-01 1.30E-14
Time (s) 4.82E+00 1.91E+00 1.88E+00 3.67E+00 3.34E+00 5.76E-01 1.08E-02 5.04E-01 4.71E-01 2.88E-01 1.74E+00 7.55E-01 5.68E-01 6.38E-01 4.73E+00 5.98E-01
F23 Best -1.05Eþ01 -1.05Eþ01 -1.01E+01 -1.05Eþ01 -8.31E+00 -1.15E+00 -3.36E+00 -1.05Eþ01 -1.05Eþ01 -1.05Eþ01 -1.05Eþ01 -1.05Eþ01 -5.13E+00 -1.05Eþ01 -1.05Eþ01 -1.05Eþ01
Worst -1.05Eþ01 -3.84E+00 -4.83E+00 -2.43E+00 -9.46E-01 -5.07E-01 -6.15E-01 -1.05E+01 -2.81E+00 -5.13E+00 -5.13E+00 -1.91E+00 -9.45E-01 -5.13E+00 -1.05Eþ01 -1.05Eþ01
Avg -1.05Eþ01 -1.00E+01 -7.24E+00 -9.06E+00 -5.09E+00 -6.05E-01 -1.06E+00 -1.05E+01 -1.02E+01 -1.04E+01 -7.19E+00 -9.73E+00 -2.28E+00 -1.02E+01 -1.05Eþ01 -1.05Eþ01
Std 1.61E-14 1.75E+00 1.54E+00 2.61E+00 5.28E-01 8.55E-02 3.59E-01 2.89E-05 1.47E+00 9.27E-01 2.64E+00 1.80E+00 1.41E+00 1.28E+00 4.25E-05 1.42E-14
Time (s) 5.41E+00 2.00E+00 2.44E+00 4.27E+00 3.42E+00 6.66E-01 1.10E-02 6.22E-01 5.63E-01 3.94E-01 1.85E+00 9.24E-01 6.67E-01 7.19E-01 5.03E+00 7.01E-01
(W|T|L) (0|10|0) (0|10|0) (0|9|1) (0|8|2) (0|10|0) (0|9|1) (0|6|4) (0|10|0) (1|8|1) (2|8|0) (0|9|1) (0|10|0) (0|10|0) (1|9|0) (0|10|0) (7|3|0)
Friedman 4.70 8.60 11.05 10.50 8.85 9.10 12.10 9.00 6.75 6.45 10.40 12.40 9.60 5.85 6.45 4.20
mean rank
Final rank 2 7 14 13 8 10 15 9 6 4 12 16 11 3 5 1
 Table 6

M. Han et al.
Results of the Wilcoxon rank-sum test on benchmark functions (F1-F23).
Function ABC-WO AOS-WO BOA-WO BWOA- ChOA- FOA-WO GA-WO GWO- MFO- PSO-WO SCSO- SHO-WO SOA-WO SSA-WO WOA-
WO WO WO WO WO WO

P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H

F1 5.64E- + 5.64E- + 5.64E- + 5.64E-39 + 5.64E-39 + 5.64E- + 5.64E- + 5.64E-39 + 5.19E-39 + 5.64E- + 5.64E-39 + 8.28E- - 5.64E- + 5.64E- + N/A =
39 39 39 39 39 39 02 39 39
F2 5.64E- + 5.64E- + 5.64E- + 5.64E-39 + 5.64E-39 + 5.64E- + 5.64E- + 5.64E-39 + 4.18E-39 + 5.64E- + 5.64E-39 + 8.28E- - 5.64E- + 5.64E- + 5.64E-39 +
39 39 39 39 39 39 02 39 39
F3 5.64E- + 5.64E- + 5.64E- + 5.64E-39 + 5.64E-39 + 5.64E- + 5.64E- + 5.64E-39 + 5.64E-39 + 5.64E- + 5.64E-39 + 5.64E- + 5.64E- + 5.64E- + 5.64E-39 +
39 39 39 39 39 39 39 39 39
F4 5.64E- + 5.64E- + 5.64E- + 5.64E-39 + 5.64E-39 + 5.64E- + 5.64E- + 5.64E-39 + 5.64E-39 + 5.64E- + 5.64E-39 + 1.23E- + 5.64E- + 5.64E- + 5.64E-39 +
39 39 39 39 39 39 03 39 39
F5 1.25E- + 1.25E- + 1.25E- + 1.25E-35 + 1.25E-35 + 1.25E- + 1.25E- + 1.25E-35 + 1.25E-35 + 1.25E- + 1.25E-35 + 3.07E- + 1.25E- + 1.25E- + 1.25E-35 +
35 35 35 35 35 35 34 35 35
F6 4.18E- + 1.25E- + 1.25E- + 1.25E-35 + 1.25E-35 + 1.25E- + 1.25E- + 1.25E-35 + 9.27E-21 + 1.25E- + 1.25E-35 + 1.25E- + 1.25E- + 2.97E- + 1.25E-35 +
24 35 35 35 35 35 35 35 14
F7 2.56E- + 4.12E- + 2.72E- + 1.33E-05 + 5.15E-22 + 2.56E- + 2.56E- + 2.31E-28 + 2.56E-34 + 2.56E- + 2.40E-07 + 8.71E- + 2.56E- + 2.56E- + 3.33E-20 +
34 32 34 34 34 34 04 34 34
F8 2.52E- + 2.52E- + 2.52E- + 2.52E-34 + 2.52E-34 + 2.52E- + 2.52E- + 2.52E-34 + 2.52E-34 + 2.52E- + 2.52E-34 + 2.32E- + 2.52E- + 2.52E- + 5.65E-33 +
34 34 34 34 34 34 25 34 34
F9 5.64E- + 1.20E- + 1.28E- + N/A = 1.33E-02 + 5.64E- + 5.64E- + N/A = 5.64E-39 + 5.64E- + 5.64E-39 = N/A = N/A = 5.63E- + N/A =
39 11 32 39 39 39 39
F10 5.64E- + 5.26E- + 2.16E- + N/A = 5.64E-39 + 5.64E- + 5.64E- + 2.32E-44 + 5.64E-39 + 5.64E- + 5.64E-39 = 4.03E- + 5.76E- + 5.48E- + 1.01E-20 +
39 38 39 39 39 39 03 45 39
F11 5.64E- + 1.33E- + N/A = N/A = 1.28E-32 + 5.64E- + 5.64E- + 4.44E-02 + 5.59E-39 + 5.64E- + 5.64E-39 = N/A = N/A = 5.64E- + 2.42E-02 +
39 02 39 39 39 39
F12 2.56E- + 2.56E- + 2.56E- + 2.56E-34 + 2.56E-34 + 2.56E- + 2.56E- + 2.56E-34 + 1.43E-01 - 2.56E- + 2.56E-34 + 2.56E- + 2.56E- + 2.21E- + 2.56E-34 +
34 34 34 34 34 34 34 34 23
15

F13 2.56E- + 2.56E- + 2.56E- + 2.56E-34 + 2.56E-34 + 2.56E- + 2.56E- + 2.56E-34 + 4.62E-01 - 2.56E- + 2.56E-34 + 2.56E- + 2.56E- + 1.98E- + 2.56E-34 +
34 34 34 34 34 34 34 34 02
F14 N/A = 3.22E- + 2.52E- + 1.43E-06 + 6.60E-32 + 3.52E- + 3.77E- + 1.05E-11 + N/A = 1.73E- + 5.01E-08 + 2.14E- + 2.16E- + N/A = 3.22E-02 +
02 28 45 39 05 15 38
F15 2.52E- + 1.80E- + 2.52E- + 9.92E-19 + 2.52E-34 + 2.52E- + 6.00E- + 1.49E-20 + 3.71E-34 + 2.41E- + 7.50E-24 + 2.52E- + 1.13E- + 3.88E- + 7.18E-34 +
34 02 34 34 34 26 34 29 19
F16 1.38E- + 4.45E- + 1.38E- + 1.23E-03 + 7.21E-26 + 1.38E- + 1.38E- + 3.51E-02 + 1.23E-03 + 1.23E- + 1.23E-03 + 3.59E- + 2.16E- + 1.23E- + 1.23E-03 +
37 03 37 37 37 03 03 06 03
F17 1.15E- + 6.99E- + 1.15E- + 1.58E-02 + 3.24E-37 + 1.15E- + 1.15E- + 3.74E-34 + 1.58E-03 + 1.58E- + 2.67E-01 - 4.72E- + 9.07E- + 1.58E- + 1.22E-02 +
38 36 38 38 38 03 36 36 01
F18 N/A = 1.10E- + 5.64E- + 3.22E-02 + 5.64E-39 + 5.64E- + 5.64E- + 2.15E-38 + N/A = N/A = 1.81E-34 + 4.07E- + 8.41E- + N/A = 1.61E-25 +
36 39 39 39 36 31
F19 N/A = 5.64E- + 5.64E- + 3.22E-02 + 5.64E-39 + 5.64E- + 5.64E- + 5.64E-39 + N/A = N/A = 1.74E-34 + 5.64E- + 5.62E- + N/A = 5.64E-39 +
39 39 39 39 39 39

Expert Systems With Applications 239 (2024) 122413

F20 5.41E- + 9.86E- + 2.21E- + 2.08E-02 + 2.74E-32 + 2.21E- + 2.21E- + 4.71E-10 + 2.33E-06 + 4.82E- - 9.62E-12 + 7.14E- + 2.21E- + 4.98E- + 1.04E-10 +
15 10 35 35 35 01 23 35 22
F21 5.64E- + 5.64E- + 5.64E- + 9.34E-06 + 5.64E-39 + 5.64E- + 5.64E- + 5.64E-39 + 1.05E-10 + 7.00E- + 5.15E-39 + 5.64E- + 5.64E- + 1.39E- + 5.64E-39 +
39 39 39 39 39 09 39 39 06
F22 5.64E- + 5.64E- + 5.64E- + 1.97E-10 + 5.64E-39 + 5.64E- + 5.64E- + 5.64E-39 + 5.08E-06 + 3.27E- + 5.27E-39 + 5.64E- + 5.64E- + 7.31E- + 5.64E-39 +
39 39 39 39 39 05 39 39 03
F23 5.64E- + 5.64E- + 5.64E- + 1.03E-07 + 5.64E-39 + 5.64E- + 5.64E- + 5.64E-39 + 4.44E-02 + 8.27E- - 5.22E-39 + 5.64E- + 5.64E- + 1.33E- + 5.63E-39 +
39 39 39 39 39 02 39 39 02
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Fig. 5. Iteration diagram of benchmark functions.

16
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Fig. 6. Boxplot of benchmark functions.

17
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Fig. 7. Sensitivity analysis of WO.

Friedman test is performed for all algorithms. The results in Fig. 8 show can complete convergence in CF2-CF4 and CF11-CF13 tests. However,
that the WO is ranked first, while SCSO, SHO and WOA are ranked most of the algorithms fall into local optima in the face of mixed and
second, third and fourth, respectively. composite function optimization problems. Compared to the statistical
results of WO most of the compared algorithms exhibit extreme insta­
bility. In general, WO can be considered as a reliable algorithm in terms
4.2. CEC 2021 benchmark suite analysis of its convergence speed and accuracy.

4.2.1. Experimental setup 4.2.4. Stability analysis

A set of single objective real-parameter numerical optimizations Stability is particularly important for the algorithms. As with the 23
from the latest Competitions on Evolutionary Computation 2021 (CEC benchmark function tests, test results were recorded and statistically
2021) was selected for testing, including 10 benchmark test functions analyzed for 100 independent runs of the CEC 2021 benchmark suite. On
and eight transformations of each function (A. W. Mohamed et al., CF1 and CF11 (Unimodal function), most of the tested algorithms
2023). CF1-CF10 is the Basic transformation functions of CEC 2021, and showed relatively stable performance, except for the DA algorithm,
CF11-CF20 is the Bias transformation functions of CEC 2021. The details which showed a high number of singular values. For the remaining 18
of the CEC 2021 are reported in Table 11. The maximum iterations and complex functions, the test results of the comparison algorithms have a
population size (search agents) are set as 500 and 100 respectively. Each high degree of dispersion. In contrast, the proposed WO still maintains a
test was run 100 times independently. high stability (see Fig. 10). When the local optimum is disturbed, the
This section considers combining the proposed WO capabilities with optimizer is still able to search for the global optimum in a stable
fifteen algorithms in the optimization field include Ali Baba and the manner. This phenomenon is consistent with the standard deviation
forty thieves (AFT) (M. Braik et al., 2022), Ant Lion Optimizer (ALO) statistics in Table 12.
(Mirjalili, 2015b), Chameleon Swarm Algorithm (CSA) (M. S. Braik,
2021), Dragonfly Algorithm (DA) (Mirjalili, 2016a), Dwarf Mongoose
Optimization Algorithm (DMOA) (Agushaka et al., 2022), Grasshopper 4.3. Experimental results on 0–1 knapsack problems
Optimization Algorithm (GOA) (Abualigah & Diabat, 2020), Moth-
Flame Optimization Algorithm (MFO) (Mirjalili, 2015a), Multi-Verse The 0–1 knapsack problem (0–1 KP) is one of the most studied NP-
Optimizer (MVO) (Mirjalili et al., 2015), Northern Goshawk Optimiza­ hard combinatorial problems in the last few decades (Lim et al.,
tion (NGO) (Dehghani et al., 2021), Simulated Annealing (SA) (Kirk­ 2016). It has practical applications in numerous areas including budget
patrick et al., 1983), Sine Cosine Algorithm (SCA) (Mirjalili, 2016b), control, telecommunication, and resource allocation, to name but a few.
Snake Optimizer (SO), Salp Swarm Algorithm (SS) (Mirjalili et al., The classical 0–1 KP can be defined as: Given a set of M items with pj
2017), Sparrow Search algorithm (SSA) (Xue & Shen, 2020), White and wj representing the profit and weight of each item j, respectively; the
Shark Optimizer (WSO) (M. Braik et al., 2022) were compared. The goal is to choose a subset of the items such that its total profit is maxi­
parameters of other compared algorithms are based on the recommen­ mized without exceeding the knapsack capacity, C. The problem can be
dations in the literature, and the parameter settings are shown in formulated as:
Table 3. ∑
M
Maximize pj xj (22)
4.2.2. Quantitative analysis j=1

Table 12 reports the optimal fitness values achieved by the selected

algorithms on the studied benchmarks suite. The P-values at the sig­ ∑
M
Subject to wj xj ≤ C, xj ∈ {0, 1}, j = 1, 2, ⋯, M (23)
nificance level α = 0.05 by Wilcoxon rank-sum test are shown in j=1
Table 13. From the results, WO outperformed the other metaheuristic
algorithms on 10 and 10 out of 10 Basic transformation functions and where, xj is a binary decision variable with xj = 1 if item j is included in
Bias transformation functions in CEC 2021 benchmarks suite, respec­ the knapsack, 0 otherwise. Without loss of generality, it may be assumed
tively. For comparison algorithms, NGO steadily found the global that all weights and profits are positive, all weights are smaller than C,
optimal in CF3, CF8, CF11, CF13, CF15 and CF18 tests, and tied for the and the overall weight of items exceeds C.
first place with WO in these tests. DA can find the global optimal in tests The original data of the simulation example is shown in Table 14.
other than CF20, but not every time, as reflected in the other three The maximum iterations and population size (search agents) are set as
statistical indicators. Friedman test and Wilcoxon rank-sum test results 500 and 100 respectively.
show that the WO can take the first rank in solving CEC 2021. The test results are shown in Table 15. Compared with GA and SA,
both WO and BPSO find the optimal solution. WO takes an average of
4.2.3. Convergence analysis 0.224849 s for 10 runs, indicating that the algorithm finds the optimal
The convergence curves of all algorithms are shown in Fig. 9. By solution from 250 feasible solutions in less than 1 s. It shows the effec­
examining these curves, the WO reflects the fastest convergence. NGO tiveness of WO in dealing with NP-hard problems.

18
 M. Han et al.
Table 7
Results of benchmark functions (F1-F13) withd = 50.
Function Comparative algorithms

ABC AOS BOA BWOA ChOA FOA GA GWO MFO PSO SCSO SHO SOA SSA WOA WO

F1 Avg 1.83E-03 1.58E-242 1.84E-03 0.00Eþ00 3.00E-40 1.89E-03 1.07E+00 5.22E-128 6.50E+03 3.84E-01 0.00Eþ00 3.94E-281 4.89E-18 1.78E-08 0.00Eþ00 0.00Eþ00
Std 2.22E-04 0.00E+00 2.27E-04 0.00Eþ00 1.68E-39 2.43E-04 1.60E-01 2.46E-127 8.33E+03 2.04E-01 0.00Eþ00 0.00E+00 3.28E-17 2.37E-09 0.00Eþ00 0.00Eþ00
F2 Avg 2.83E+00 9.92E-127 2.85E+00 0.00Eþ00 7.14E-02 2.83E+00 4.74E+00 2.32E-74 6.61E+01 1.09E+00 3.61E-250 7.26E-167 2.26E-18 7.81E-01 3.00E-222 0.00Eþ00
Std 1.45E-01 4.06E-126 1.40E-01 0.00Eþ00 7.14E-01 1.36E-01 4.38E-01 3.89E-74 3.53E+01 4.04E-01 0.00E+00 0.00E+00 5.52E-18 8.14E-01 0.00E+00 0.00Eþ00
F3 Avg 1.33E+00 2.00E-182 1.34E+00 0.00Eþ00 6.66E-04 1.34E+00 6.21E+02 6.68E-29 3.74E+04 2.99E+02 0.00Eþ00 5.42E+04 9.27E+02 3.34E+01 1.33E+04 0.00Eþ00
Std 1.43E-01 0.00E+00 1.46E-01 0.00Eþ00 6.64E-03 1.55E-01 7.49E+02 6.07E-28 2.43E+04 6.25E+01 0.00Eþ00 3.13E+04 1.64E+03 2.67E+01 9.38E+03 0.00Eþ00
F4 Avg 1.01E-02 8.45E-118 1.01E-02 0.00Eþ00 1.18E-08 1.02E-02 4.96E-01 5.01E-29 6.59E+01 2.25E+00 3.25E-206 2.25E-29 5.14E-01 5.65E+00 3.02E+01 0.00Eþ00
Std 7.71E-04 4.54E-117 8.18E-04 0.00Eþ00 5.46E-08 7.59E-04 3.00E-02 8.76E-29 6.76E+00 2.23E-01 0.00E+00 2.25E-28 2.03E+00 1.94E+00 3.16E+01 0.00Eþ00
F5 Avg 4.90E+01 4.53E+01 4.90E+01 4.89E+01 4.88E+01 4.90E+01 1.63E+02 4.62E+01 3.22E+06 5.65E+02 4.70E+01 1.16E+01 4.82E+01 8.96E+01 4.44E+01 1.30E-03
Std 1.97E-01 4.57E-01 1.71E-01 4.14E-02 2.00E-01 1.64E-01 2.15E+01 8.60E-01 1.58E+07 4.29E+02 1.07E+00 2.03E+01 5.91E-01 9.69E+01 6.38E+00 3.20E-03
F6 Avg 1.28E+01 2.94E-03 1.28E+01 9.90E+00 5.58E+00 1.28E+01 1.86E+01 1.03E+00 6.40E+03 3.84E-01 1.39E+00 4.18E-01 2.77E+00 9.41E-07 4.01E-04 1.73E-08
Std 1.45E-02 9.28E-04 1.45E-02 1.12E+00 3.97E-01 1.42E-02 6.70E-01 4.35E-01 8.70E+03 2.11E-01 5.07E-01 6.10E-01 5.76E-01 1.32E-06 1.30E-04 2.32E-09
19

F7 Avg 8.61E+01 1.69E-04 8.99E+01 2.42E-05 1.96E-04 9.02E+01 3.81E+00 1.86E-04 1.38E+01 5.10E+01 1.21E-05 2.45E-05 4.88E-02 4.35E-02 2.81E-04 1.16E-05
Std 2.41E+01 8.63E-05 2.34E+01 2.44E-05 1.50E-04 2.87E+01 7.90E-01 1.19E-04 2.11E+01 5.55E+01 1.44E-05 4.44E-05 1.90E-02 1.28E-02 3.28E-04 9.45E-06
F8 Avg -3.58E+00 -1.20E+04 -2.81E+00 -7.41E+03 -9.41E+03 -3.44E+00 -3.37E+03 -9.77E+03 -1.37E+04 -1.11E+04 -1.16E+04 -2.10E+04 -8.21E+03 -9.55E+04 -2.03E+04 -2.10Eþ04
Std 5.94E+00 8.64E+02 8.40E-01 9.04E+02 6.99E+01 6.03E+00 6.62E+02 9.75E+02 1.60E+03 1.23E+03 1.01E+03 1.19E+01 4.29E+02 1.70E+04 1.32E+03 1.60E-02
F9 Avg 1.24E+02 5.80E+00 1.25E+02 0.00Eþ00 2.19E-01 1.25E+02 6.44E+01 5.68E-16 2.50E+02 1.55E+02 0.00Eþ00 0.00Eþ00 0.00Eþ00 6.50E+01 0.00Eþ00 0.00Eþ00
Std 1.28E+01 1.83E+01 1.19E+01 0.00Eþ00 1.33E+00 1.30E+01 5.96E+00 5.68E-15 5.59E+01 3.26E+01 0.00Eþ00 0.00Eþ00 0.00Eþ00 2.18E+01 0.00Eþ00 0.00Eþ00
F10 Avg 9.45E-02 4.41E-15 9.50E-02 8.88E-16 2.00E+01 9.40E-02 1.09E+00 1.42E-14 1.78E+01 1.06E+00 8.88E-16 1.03E-15 8.10E-11 2.05E+00 3.27E-15 8.88E-16
Std 6.27E-03 3.55E-16 6.38E-03 0.00Eþ00 9.73E-04 6.14E-03 1.25E-01 2.04E-15 5.40E+00 5.62E-01 0.00Eþ00 7.00E-16 2.95E-10 6.59E-01 2.43E-15 0.00Eþ00
F11 Avg 1.93E-06 7.68E-04 1.92E-06 0.00Eþ00 1.44E-03 1.94E-06 2.62E-02 2.27E-04 4.97E+01 1.42E-02 0.00Eþ00 0.00Eþ00 1.50E-16 4.24E-03 4.71E-04 0.00Eþ00
Std 3.34E-07 3.28E-03 2.87E-07 0.00Eþ00 6.06E-03 3.56E-07 3.82E-03 1.61E-03 6.47E+01 7.59E-03 0.00Eþ00 0.00Eþ00 9.16E-16 6.17E-03 2.69E-03 0.00Eþ00
F12 Avg 1.52E+00 2.74E-03 1.52E+00 6.23E-01 3.81E-01 1.52E+00 1.90E+00 4.23E-02 1.28E+07 9.07E-03 4.24E-02 1.39E-02 1.53E-01 4.11E+00 2.19E-05 4.67E-09
Std 3.05E-03 1.25E-02 2.75E-03 1.60E-01 1.21E-01 3.01E-03 5.35E-02 3.36E-02 5.61E+07 2.03E-02 1.79E-02 1.59E-02 5.42E-02 2.05E+00 7.68E-06 5.04E-09
F13 Avg 4.81E+00 5.29E-02 4.81E+00 4.98E+00 4.86E+00 4.81E+00 1.58E-01 7.89E-01 1.64E+07 1.35E-01 4.23E+00 3.43E-01 4.68E+00 7.42E-01 3.84E-03 2.27E-07
Std 2.00E-02 1.53E-01 1.72E-02 1.27E-01 1.26E-01 1.69E-02 3.10E-02 2.64E-01 8.08E+07 6.29E-02 3.83E-01 4.22E-01 8.46E-02 5.66E+00 5.36E-03 2.93E-07

Expert Systems With Applications 239 (2024) 122413

(W|T|L) (0|13|0) (0|13|0) (0|12|1) (7|6|0) (0|12|1) (0|12|1) (0|13|0) (0|13|0) (0|4|9) (0|13|0) (5|8|0) (2|10|1) (0|13|0) (0|13|0) (2|11|0) (13|0|0)
Friedman mean rank 11.15 5.46 11.69 5.65 9.54 11.77 12.54 6.85 14.69 10.69 4.42 5.12 8.73 9.85 5.85 2.00
Final rank 12 4 13 5 9 14 15 7 16 11 2 3 8 10 6 1
 M. Han et al.
Table 8
Results of benchmark functions (F1-F13) withd = 100.
.Function Comparative algorithms

ABC AOS BOA BWOA ChOA FOA GA GWO MFO PSO SCSO SHO SOA SSA WOA WO

F1 Avg 4.53E-03 7.97E-232 4.49E-03 0.00Eþ00 4.04E-24 4.47E-03 2.09E+01 2.43E-86 1.75E+04 2.50E+01 0.00Eþ00 1.76E-247 3.28E+02 1.07E-07 0.00Eþ00 0.00Eþ00
Std 3.48E-04 0.00E+00 3.83E-04 0.00Eþ00 2.68E-23 4.01E-04 6.44E-01 3.16E-86 1.24E+04 5.72E+00 0.00Eþ00 0.00E+00 7.06E+02 1.35E-08 0.00Eþ00 0.00Eþ00
F2 Avg 6.20E+00 1.54E-122 6.15E+00 0.00Eþ00 3.08E-17 6.17E+00 3.75E+01 6.58E-51 1.42E+02 3.17E+01 2.39E-241 4.86E-152 6.11E-04 3.94E+00 5.82E-224 0.00Eþ00
Std 2.22E-01 6.22E-122 2.63E-01 0.00Eþ00 5.07E-17 2.61E-01 7.20E-01 5.15E-51 4.99E+01 7.58E+00 0.00E+00 4.86E-151 7.26E-04 1.80E+00 0.00E+00 0.00Eþ00
F3 Avg 1.22E+01 2.44E-173 1.25E+01 0.00Eþ00 9.82E+01 1.25E+01 8.78E+03 2.07E-10 1.33E+05 5.90E+03 0.00Eþ00 5.55E+05 1.10E+05 4.39E+03 2.71E+05 0.00Eþ00
Std 9.97E-01 0.00E+00 1.06E+00 0.00Eþ00 3.62E+02 9.92E-01 2.26E+03 1.15E-09 5.28E+04 1.14E+03 0.00Eþ00 1.28E+05 3.01E+04 3.17E+03 7.34E+04 0.00Eþ00
F4 Avg 1.35E-02 1.08E-113 1.35E-02 0.00Eþ00 1.50E+01 1.34E-02 8.86E-01 2.13E-13 8.91E+01 5.50E+00 3.30E-206 9.94E-06 2.83E+01 1.35E+01 5.92E+01 0.00Eþ00
Std 9.58E-04 5.56E-113 1.05E-03 0.00Eþ00 2.79E+01 8.60E-04 1.06E-02 1.39E-12 2.74E+00 5.69E-01 0.00E+00 9.94E-05 8.95E+00 1.76E+00 3.34E+01 0.00Eþ00
F5 Avg 9.96E+01 9.59E+01 9.96E+01 9.89E+01 9.88E+01 9.97E+01 1.20E+03 9.65E+01 4.40E+07 1.77E+04 9.76E+01 1.02E+01 2.39E+07 1.79E+02 9.45E+01 4.69E-03
Std 2.21E-01 7.90E-01 2.51E-01 4.31E-02 1.42E-01 2.72E-01 3.30E+01 9.49E-01 6.04E+07 5.34E+03 9.34E-01 2.94E+01 1.61E+07 1.96E+02 9.38E+00 1.07E-02
F6 Avg 2.56E+01 1.64E-01 2.56E+01 2.22E+01 1.62E+01 2.56E+01 8.37E+01 6.17E+00 1.73E+04 2.47E+01 6.18E+00 6.89E-01 3.10E+02 5.16E-05 1.18E-02 1.05E-07
Std 2.50E-02 6.34E-02 2.51E-02 1.25E+00 6.51E-01 2.31E-02 1.10E+00 8.06E-01 1.22E+04 7.02E+00 9.87E-01 1.17E+00 4.67E+02 7.35E-05 2.68E-03 1.28E-08
20

F7 Avg 6.08E+02 2.34E-04 6.07E+02 2.48E-05 3.51E-04 6.29E+02 4.30E+02 3.25E-04 9.66E+01 1.22E+03 1.17E-05 1.91E-05 2.08E+01 1.94E-01 2.44E-04 1.43E-05
Std 1.33E+02 1.37E-04 1.45E+02 2.44E-05 2.36E-04 1.37E+02 2.20E+01 1.79E-04 8.94E+01 2.74E+02 1.43E-05 2.23E-05 1.69E+01 4.44E-02 2.43E-04 1.13E-05
F8 Avg -3.59E+00 -1.96E+04 -3.45E+00 -1.09E+04 -1.83E+04 -4.31E+00 -4.82E+03 -1.70E+04 -2.49E+04 -2.13E+04 -2.17E+04 -4.19E+04 -1.46E+04 -1.75E+05 -4.09E+04 -4.19Eþ04
Std 2.05E+00 1.58E+03 5.52E-01 1.58E+03 8.63E+01 6.61E+00 8.55E+02 2.19E+03 2.90E+03 2.22E+03 1.68E+03 1.78E+02 7.54E+02 2.77E+04 1.81E+03 3.13E-01
F9 Avg 2.97E+02 1.25E+00 2.92E+02 0.00Eþ00 7.43E-02 2.94E+02 7.28E+02 2.27E-15 5.80E+02 6.87E+02 0.00Eþ00 0.00Eþ00 1.30E+02 1.17E+02 2.27E-15 0.00Eþ00
Std 1.86E+01 1.25E+01 2.07E+01 0.00Eþ00 7.43E-01 1.86E+01 1.78E+01 1.60E-14 8.40E+01 1.10E+02 0.00Eþ00 0.00Eþ00 5.19E+01 3.58E+01 1.60E-14 0.00Eþ00
F10 Avg 1.07E-01 4.55E-15 1.07E-01 8.88E-16 2.00E+01 1.06E-01 3.38E+00 2.43E-14 1.98E+01 3.71E+00 8.88E-16 1.17E-15 1.30E+00 3.20E+00 4.12E-15 8.88E-16
Std 6.11E-03 7.91E-16 5.38E-03 0.00Eþ00 3.50E-04 5.65E-03 3.57E-02 3.39E-15 3.27E-01 3.10E-01 0.00Eþ00 1.09E-15 1.54E+00 6.03E-01 2.43E-15 0.00Eþ00
F11 Avg 2.62E-06 9.87E-05 2.63E-06 0.00Eþ00 2.13E-03 2.63E-06 2.95E-01 2.06E-04 2.12E+02 2.97E-01 0.00Eþ00 0.00Eþ00 3.35E+00 3.45E-03 0.00Eþ00 0.00Eþ00
Std 3.36E-07 9.87E-04 3.22E-07 0.00Eþ00 7.81E-03 3.65E-07 1.23E-02 1.45E-03 1.30E+02 6.71E-02 0.00Eþ00 0.00Eþ00 4.26E+00 5.39E-03 0.00Eþ00 0.00Eþ00
F12 Avg 1.38E+00 1.23E-02 1.38E+00 8.62E-01 6.43E-01 1.38E+00 3.46E+00 1.23E-01 4.61E+07 6.01E-01 9.18E-02 1.60E-02 6.91E+07 6.57E+00 1.56E-04 6.04E-08
Std 1.98E-03 1.18E-02 2.27E-03 1.13E-01 7.83E-02 2.10E-03 5.73E-02 3.43E-02 1.05E+08 3.65E-01 2.22E-02 3.14E-02 5.01E+07 1.54E+00 4.11E-05 7.43E-08
F13 Avg 9.77E+00 6.16E+00 9.77E+00 9.99E+00 9.66E+00 9.77E+00 3.12E+00 4.52E+00 1.60E+08 1.56E+01 9.45E+00 7.00E-01 1.25E+08 1.15E+02 3.10E-02 4.32E-06
Std 1.91E-02 2.78E+00 2.07E-02 4.92E-03 1.46E-01 2.33E-02 9.10E-02 4.47E-01 2.60E+08 6.30E+00 1.54E-01 9.82E-01 8.38E+07 2.64E+01 2.56E-02 6.47E-06

Expert Systems With Applications 239 (2024) 122413

(W|T|L) (0|13|0) (0|13|0) (0|12|1) (7|6|0) (0|12|1) (0|13|0) (0|12|1) (0|13|0) (0|6|7) (0|12|1) (5|8|0) (2|10|1) (0|12|1) (0|13|0) (2|11|0) (13|0|0)
Friedman mean rank 10.73 5.62 10.58 5.31 9.23 10.54 12.39 6.58 14.31 12.31 3.85 4.69 13.00 9.69 5.39 1.81
Final rank 12 6 11 4 8 10 14 7 16 13 2 3 15 9 5 1
 M. Han et al.
Table 9
Results of benchmark functions (F1-F13) with d = 500.
Function Comparative algorithms

ABC AOS BOA BWOA ChOA FOA GA GWO MFO PSO SCSO SHO SOA SSA WOA WO

F1 Avg 3.43E-02 3.16E-225 3.42E-02 0.00Eþ00 8.50E-09 3.41E-02 1.36E+02 4.44E-38 6.24E+05 3.68E+03 0.00Eþ00 4.78E-262 4.90E+05 1.50E+03 0.00Eþ00 0.00Eþ00
Std 1.89E-03 0.00E+00 1.51E-03 0.00Eþ00 1.12E-08 1.86E-03 1.46E+00 3.30E-38 3.26E+04 2.37E+02 0.00Eþ00 0.00E+00 3.75E+04 2.08E+02 0.00Eþ00 0.00Eþ00
F2 Avg 3.55E+01 4.95E-118 3.55E+01 0.00Eþ00 1.21E-06 3.54E+01 2.20E+02 5.14E-23 1.63E+03 4.63E+107 1.07E-228 1.13E-148 4.82E+02 1.43E+02 9.00E-221 0.00Eþ00
Std 7.61E-01 1.33E-117 7.20E-01 0.00Eþ00 6.58E-07 6.64E-01 1.12E+00 1.66E-23 7.43E+01 4.59E+108 0.00E+00 1.13E-147 5.24E+01 9.70E+00 0.00E+00 0.00Eþ00
F3 Avg 2.06E+03 1.83E-158 2.06E+03 0.00Eþ00 9.39E+05 2.06E+03 2.66E+05 5.53E+03 2.47E+06 3.10E+05 0.00Eþ00 2.24E+07 2.82E+07 2.61E+05 2.13E+07 0.00Eþ00
Std 9.25E+01 1.41E-157 9.47E+01 0.00Eþ00 5.43E+05 8.40E+01 5.08E+04 6.08E+03 5.18E+05 6.45E+04 0.00Eþ00 6.08E+06 6.64E+06 1.51E+05 4.35E+06 0.00Eþ00
F4 Avg 2.83E-02 1.16E-108 2.82E-02 0.00Eþ00 9.74E+01 2.85E-02 9.72E-01 2.89E+01 9.85E+01 2.38E+01 1.40E-201 8.30E-34 7.80E+01 2.37E+01 7.43E+01 0.00Eþ00
Std 1.72E-03 6.22E-108 1.48E-03 0.00Eþ00 1.15E+00 1.51E-03 2.91E-03 8.08E+00 3.44E-01 1.08E+00 0.00E+00 8.30E-33 1.74E+01 1.55E+00 2.78E+01 0.00Eþ00
F5 Avg 5.12E+02 4.97E+02 5.12E+02 4.99E+02 4.99E+02 5.12E+02 7.98E+03 4.97E+02 2.32E+09 1.64E+07 4.98E+02 1.37E+01 3.56E+09 8.21E+04 4.94E+02 5.12E-02
Std 1.36E+00 4.28E-01 1.26E+00 4.18E-02 1.87E-01 1.38E+00 1.06E+02 3.84E-01 2.10E+08 2.14E+06 3.66E-01 6.95E+01 2.59E+08 1.87E+04 3.81E-01 1.64E-01
F6 Avg 1.29E+02 3.74E+01 1.29E+02 1.22E+02 1.12E+02 1.29E+02 4.81E+02 8.43E+01 6.27E+05 3.71E+03 7.64E+01 5.74E+00 4.89E+05 1.55E+03 1.49E+00 2.03E-03
Std 7.84E-02 2.11E+00 6.33E-02 1.35E+00 1.03E+00 7.01E-02 2.63E+00 2.21E+00 3.47E+04 2.74E+02 4.02E+00 9.41E+00 4.29E+04 2.15E+02 2.33E-01 4.14E-03
21

F7 Avg 6.44E+04 2.46E-04 6.36E+04 1.21E-05 2.05E-03 6.14E+04 1.73E+04 1.18E-03 1.74E+04 5.46E+04 2.64E-05 1.46E-05 2.51E+04 7.57E+00 3.58E-04 2.19E-05
Std 1.00E+04 1.60E-04 8.71E+03 1.14E-05 1.67E-03 9.28E+03 2.76E+02 3.96E-04 1.91E+03 2.09E+03 3.45E-05 1.39E-05 1.84E+03 7.75E-01 4.22E-04 2.87E-05
F8 Avg -4.98E+00 -5.03E+04 -4.61E+00 -2.58E+04 -8.66E+04 -4.71E+00 -1.33E+04 -6.63E+04 -8.88E+04 -1.05E+05 -9.02E+04 -2.09E+05 -5.49E+04 -8.00E+04 -2.06E+05 -2.10Eþ05
Std 7.41E+00 6.97E+03 4.40E+00 3.73E+03 5.27E+02 4.81E+00 2.22E+03 1.14E+04 7.96E+03 1.05E+04 5.27E+03 2.79E+02 2.38E+03 1.05E+04 6.68E+03 2.65Eþ01
F9 Avg 1.83E+03 0.00Eþ00 1.84E+03 0.00Eþ00 2.16E-02 1.85E+03 4.42E+03 1.22E-12 5.37E+03 7.18E+03 0.00Eþ00 0.00Eþ00 2.23E+03 6.57E+02 0.00Eþ00 0.00Eþ00
Std 4.74E+01 0.00Eþ00 6.27E+01 0.00Eþ00 2.15E-01 5.54E+01 3.62E+01 5.52E-13 1.52E+02 4.09E+02 0.00Eþ00 0.00Eþ00 2.57E+02 9.29E+01 0.00Eþ00 0.00Eþ00
F10 Avg 1.37E-01 4.55E-15 1.38E-01 8.88E-16 2.00E+01 1.38E-01 3.67E+00 9.16E-14 2.00E+01 1.06E+01 8.88E-16 1.10E-15 1.82E+01 9.75E+00 3.48E-15 8.88E-16
Std 4.58E-03 6.09E-16 3.83E-03 0.00Eþ00 8.25E-03 4.13E-03 1.16E-02 8.02E-15 2.44E-02 2.32E-01 0.00Eþ00 9.87E-16 3.43E-01 4.11E-01 2.57E-15 0.00Eþ00
F11 Avg 4.99E-06 0.00Eþ00 4.93E-06 0.00Eþ00 2.27E-03 4.96E-06 4.78E-01 4.78E-04 5.65E+03 1.95E+00 0.00Eþ00 0.00Eþ00 4.43E+03 1.49E+01 0.00Eþ00 0.00Eþ00
Std 4.33E-07 0.00Eþ00 4.02E-07 0.00Eþ00 1.27E-02 4.08E-07 9.45E-03 2.76E-03 3.07E+02 6.12E-02 0.00Eþ00 0.00Eþ00 3.65E+02 1.81E+00 0.00Eþ00 0.00Eþ00
F12 Avg 1.26E+00 1.37E-01 1.26E+00 1.12E+00 1.04E+00 1.26E+00 3.78E+00 6.23E-01 5.13E+09 5.67E+04 3.63E-01 1.69E-02 1.07E+10 1.52E+01 2.13E-03 3.26E-07
Std 1.21E-03 1.23E-02 1.19E-03 3.71E-02 2.42E-02 1.15E-03 2.50E-02 2.83E-02 6.28E+08 2.53E+04 3.46E-02 2.93E-02 6.81E+08 3.21E+00 3.22E-04 6.29E-07
F13 Avg 4.96E+01 4.98E+01 4.96E+01 5.00E+01 4.89E+01 4.96E+01 2.02E+01 4.38E+01 9.87E+09 1.12E+06 4.96E+01 3.15E+00 1.73E+10 9.15E+02 8.63E-01 1.32E-04
Std 3.46E-02 1.94E-02 3.91E-02 4.96E-03 2.15E-01 4.13E-02 2.23E-01 6.15E-01 1.12E+09 3.10E+05 8.38E-02 5.02E+00 1.31E+09 4.32E+01 2.13E-01 3.02E-04

Expert Systems With Applications 239 (2024) 122413

(W|T|L) (0|12|1) (0|11|2) (0|12|1) (7|6|0) (0|12|1) (0|13|0) (0|13|0) (0|13|0) (0|9|4) (0|11|2) (3|10|0) (2|11|0) (0|9|4 (0|13|0) (3|10|0) (13|0|0)
Friedman mean rank 10.12 5.54 10.04 5.04 9.31 9.65 11.08 7.39 14.23 12.92 4.15 4.39 14.31 10.85 4.96 2.04
Final rank 11 6 10 5 8 9 13 7 15 14 2 3 16 12 4 1
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Table 10
Average fitness values of WO using different parameters.
Parameter Value F5 F6 F12 F13 F14 F15

T 200 2.25E-04 3.32E-08 2.98E-10 1.23E-08 9.98E-01 3.08E-04

500 8.82E-04 5.91E-08 5.97E-10 2.01E-08 9.98E-01 3.08E-04
1000 1.62E-04 6.56E-08 1.27E-09 6.82E-09 9.98E-01 3.08E-04
2000 1.48E-05 2.14E-08 2.05E-10 1.97E-10 9.98E-01 3.07E-04
N 30 8.42E-03 1.45E-05 1.12E-07 1.35E-06 9.98E-01 3.14E-04
50 6.64E-04 5.24E-07 1.73E-09 1.03E-07 9.98E-01 3.09E-04
80 4.89E-04 1.58E-07 4.37E-10 1.53E-08 9.98E-01 3.08E-04
100 2.17E-05 2.83E-08 2.78E-10 1.87E-10 9.98E-01 3.07E-04
p 0.3 1.40E-02 6.10E-04 3.51E-06 1.66E-05 9.98E-01 3.82E-04
0.35 9.17E-04 3.19E-05 1.12E-07 1.35E-07 9.98E-01 3.12E-04
0.4 2.86E-04 5.58E-07 5.00E-09 1.56E-08 9.98E-01 3.08E-04
0.45 1.73E-05 2.83E-08 3.14E-10 4.11E-10 9.98E-01 3.07E-04

Fig. 8. Friedman mean rank results for different dimensions.

Table 11 agents) are set as 500 and 100 respectively.

Summary of the CEC 2021 benchmark suite.
Type No. Functions fmin 4.4.1. Case 1: Gear train design
Basic Bias The gear train design problem refers to the cost optimization of the
Unimodal function CF1 CEC 2017 F1 (Awad et al., 0 100 gear ratio of a compound gear train (see Fig. 11(a)). The current gear
2017) design problem has four gears and the error between a required gear
Basic functions CF2 CEC 2014 F11 (Liang et al., 0 1100 ratio (1/6.931) and an obtained gear ratio are shown to be minimized.
2014) Therefore, the objective function is considered a discrete problem with
CF3 CEC 2017 F7 0 700
CF4 CEC 2017 F19 0 1900
error reduction. The problem variables are the teeth for each gearwheel.
Hybrid functions CF5 CEC 2014 F17 0 1700 The design engineering constraint is defined as the number of teeth on
CF6 CEC 2017 F16 0 1600 any gear that should only be in the range of [12, 60]. To handle discrete
CF7 CEC 2014 F21 0 2100 variables, each search agent was rounded to the nearest integer number
Composition CF8 CEC 2017 F22 0 2200
before the fitness evaluation (Mirjalili, 2015b; Peraza-Vázquez et al.,
functions CF9 CEC 2017 F24 0 2400
CF10 CEC 2017 F25 0 2500 2021). The mathematical model of pressure vessel design is as follows:
# For all functions: search range is [100, 100]d Suppose x = [x1 x2 x3 x4 ] = [GA GB GC GD ]
( )2
1 x3 x2
Minimize f (x) = − (24)
4.4. Experimental results on real-world engineering optimization problems 6.931 x1 x4

The proposed WO was also employed to solve 6 different engineering where x1,2,3,4 ∈ [12, 60] are integer design variables.
design problems and the results are presented in this section (Gandomi, Table 16 shows the values of the four design variables and the
2014). The engineering optimization problems used in this study find optimal cost. From the table, it can be observed that the WO provides
the optimal solutions under special conditions, such as design principles, better weight than other methods and is suitable to solve discrete con­
resource limitations, and safety requirements (Han et al., 2022). Since strained problems, followed by Cuckoo Search (CS).
the problems of this section have different constraints, we need to
employ a constraint handling method. There are different types of 4.4.2. Case 2: Cantilever beam design
penalty functions in the literature (Coello Coello, 2002): static, dynamic, The design of cantilever beam is a type of concrete engineering
annealing, adaptive, co-evolutionary, and death penalty. The last pen­ optimization problems (see Fig. 11(b)). The weight of the cantilever
alty function, death penalty, is the simplest method that assigns a large beam is minimized by optimizing five hollow square section parameters.
objective value to the fitness function (in the case of minimization). This The mathematical model of cantilever beam design is as follows:
approach causes the algorithm to discard solutions that are much larger Suppose x = [x1 x2 x3 x4 x5 ]
than the normal range during the optimization process. For the sake of Minimize f (x) = 0.06224(x1 + x2 + x3 + x4 + x5 ) (25)
simplicity, this section equips the WO with a death penalty function to
handle constraints. The maximum iterations and population size (search

22
 M. Han et al.
Table 12
Results of the comparative methods on CEC 2021 benchmark suite.
Function AFT ALO CSA DA DMOA GOA MFO MVO NGO SA SCA SO SS SSA WSO WO

CF1 Best 1.57E-31 1.60E-02 2.81E-02 0.00Eþ00 4.89E-19 3.41E-01 1.03E-11 1.23E+03 8.09E-99 6.06E-18 2.08E-20 5.80E-125 1.40E-02 2.17E-01 4.89E-11 3.39E-279
Worst 4.36E-27 5.94E+03 3.02E+03 1.36E+08 7.79E-17 9.84E+03 1.00E+04 1.62E+04 2.48E-95 1.40E-06 1.31E-10 1.01E-116 9.92E+03 4.71E+03 3.87E-02 8.68E-212
Avg 2.21E-28 9.26E+02 3.93E+02 4.89E+06 1.33E-17 2.01E+03 3.50E+03 6.84E+03 1.65E-96 1.41E-08 2.85E-12 3.73E-118 6.84E+02 6.97E+02 7.27E-04 8.69E-214
Std 5.08E-28 1.10E+03 5.54E+02 1.70E+07 1.25E-17 2.58E+03 4.79E+03 3.65E+03 3.00E-96 1.40E-07 1.37E-11 1.55E-117 1.26E+03 9.64E+02 4.14E-03 0.00Eþ00
CF2 Best 3.75E-01 3.67E+00 1.25E-01 0.00Eþ00 7.61E+02 3.79E+00 6.83E+00 3.68E+00 0.00Eþ00 6.25E-02 0.00Eþ00 0.00Eþ00 5.40E-09 1.25E-01 6.29E-02 0.00Eþ00
Worst 8.37E+02 1.37E+03 7.41E+02 1.56E+03 1.38E+03 1.68E+03 1.52E+03 1.30E+03 9.40E-10 1.37E+01 9.40E+02 2.64E+01 1.28E+03 9.50E+02 2.79E+02 0.00Eþ00
Avg 2.76E+02 7.46E+02 1.40E+02 5.34E+02 1.10E+03 9.10E+02 4.70E+02 5.05E+02 1.19E-11 5.24E+00 1.90E+01 4.87E+00 4.23E+02 4.08E+02 3.90E+01 0.00Eþ00
Std 1.93E+02 3.00E+02 1.77E+02 4.34E+02 1.45E+02 3.44E+02 3.53E+02 2.79E+02 9.70E-11 3.65E+00 1.01E+02 5.49E+00 2.72E+02 2.81E+02 7.14E+01 0.00Eþ00
CF3 Best 4.98E+00 8.96E+00 3.98E+00 0.00Eþ00 1.63E+01 9.95E+00 4.98E+00 7.99E+00 0.00Eþ00 3.65E-17 6.66E-21 7.89E-31 3.98E+00 4.98E+00 3.84E-01 0.00Eþ00
Worst 4.80E+01 8.10E+01 2.93E+01 1.98E+02 3.82E+01 6.03E+01 4.15E+01 4.22E+01 0.00Eþ00 1.30E+01 9.68E+01 1.61E+01 5.77E+01 5.67E+01 4.06E+01 0.00Eþ00
Avg 2.17E+01 3.12E+01 1.56E+01 3.05E+01 3.08E+01 2.61E+01 2.25E+01 2.51E+01 0.00Eþ00 1.07E+01 1.12E+01 7.79E+00 2.42E+01 2.33E+01 2.16E+01 0.00Eþ00
Std 8.02E+00 1.16E+01 4.65E+00 2.48E+01 3.74E+00 1.13E+01 6.84E+00 6.95E+00 0.00Eþ00 2.18E+00 2.81E+01 5.15E+00 9.98E+00 1.10E+01 7.57E+00 0.00Eþ00
CF4 Best 2.51E-01 3.03E-01 3.30E-01 0.00Eþ00 1.12E+00 3.59E-01 2.17E-01 2.62E-01 0.00Eþ00 6.52E-02 0.00Eþ00 0.00Eþ00 3.20E-01 3.06E-01 8.22E-01 0.00Eþ00
Worst 2.49E+00 3.80E+00 2.38E+00 1.02E+01 2.93E+00 3.63E+00 3.03E+00 3.48E+00 8.54E-02 1.75E+00 5.50E+00 1.28E-01 5.03E+00 3.86E+00 2.45E+00 0.00Eþ00
Avg 9.57E-01 1.29E+00 9.60E-01 1.99E+00 2.12E+00 1.53E+00 1.10E+00 1.32E+00 1.93E-03 6.61E-01 5.18E-01 3.50E-03 1.47E+00 1.43E+00 1.83E+00 0.00Eþ00
Std 3.97E-01 6.31E-01 4.13E-01 1.68E+00 3.56E-01 6.79E-01 5.64E-01 5.57E-01 1.12E-02 3.05E-01 1.20E+00 1.96E-02 8.15E-01 7.24E-01 3.19E-01 0.00Eþ00
CF5 Best 4.16E-01 3.36E+02 7.43E+00 0.00Eþ00 5.43E+00 6.41E+02 1.20E+00 1.63E+02 9.19E-22 8.49E-20 4.81E-19 2.50E-117 2.72E+02 2.92E+02 2.72E-12 3.10E-295
Worst 4.93E+02 1.13E+04 5.00E+02 1.33E+05 4.52E+01 1.35E+04 1.71E+03 1.38E+03 1.34E-19 2.26E+01 2.61E+01 2.85E+02 6.57E+03 1.09E+04 6.76E+00 4.78E-243
Avg 1.14E+02 3.96E+03 1.22E+02 5.29E+03 1.65E+01 5.37E+03 1.32E+02 6.58E+02 1.65E-20 2.49E+00 4.78E-01 4.03E+01 2.12E+03 2.07E+03 1.29E+00 4.79E-245
Std 1.07E+02 2.83E+03 1.27E+02 1.43E+04 7.54E+00 3.54E+03 2.41E+02 2.78E+02 1.66E-20 4.62E+00 3.11E+00 6.68E+01 1.46E+03 1.73E+03 1.12E+00 0.00Eþ00
CF6 Best 2.38E-01 2.29E+00 8.38E-01 0.00Eþ00 1.25E+00 2.27E+00 3.63E-01 1.53E+00 1.46E-03 2.29E-02 2.17E-04 1.13E-14 6.53E-01 4.68E-01 1.78E-01 0.00Eþ00
Worst 1.41E+02 2.53E+02 1.51E+01 3.61E+02 5.86E+00 2.77E+02 1.72E+02 2.43E+02 8.34E-02 1.24E+01 1.84E+01 1.19E+01 1.20E+02 1.61E+02 1.16E+01 0.00Eþ00
Avg 1.08E+01 3.11E+01 2.86E+00 5.44E+01 2.45E+00 8.85E+01 2.53E+01 3.92E+01 1.84E-02 2.46E+00 4.49E-01 9.29E-01 7.92E+00 1.23E+01 3.29E+00 0.00Eþ00
Std 1.91E+01 4.08E+01 1.88E+00 7.88E+01 6.41E-01 7.41E+01 3.36E+01 5.35E+01 1.10E-02 4.11E+00 1.90E+00 1.35E+00 1.33E+01 2.21E+01 2.59E+00 0.00Eþ00
CF7 Best 4.57E-02 5.69E+01 8.05E-01 0.00Eþ00 9.58E-01 4.25E+02 1.42E-02 1.46E+01 6.42E-05 1.19E-02 7.35E-06 6.66E-08 1.52E+01 1.02E+02 2.91E-02 5.06E-302
23

Worst 2.50E+02 1.29E+04 3.37E+02 1.13E+04 6.34E+00 1.23E+04 4.43E+02 6.23E+02 1.80E-03 3.40E+01 3.92E-01 1.36E+02 9.10E+03 6.02E+03 2.97E+00 1.14E-241
Avg 4.42E+01 3.56E+03 3.59E+01 1.58E+03 1.80E+00 4.81E+03 3.55E+01 1.53E+02 5.09E-04 3.12E+00 4.82E-02 5.90E+00 1.57E+03 1.21E+03 4.97E-01 1.39E-243
Std 6.10E+01 3.18E+03 6.37E+01 2.77E+03 7.50E-01 3.19E+03 6.92E+01 1.23E+02 2.82E-04 6.43E+00 7.40E-02 1.96E+01 1.88E+03 1.12E+03 4.62E-01 0.00Eþ00
CF8 Best 0.00Eþ00 1.04E+01 1.11E-15 0.00Eþ00 1.14E-02 4.91E-05 1.53E-13 9.97E+00 0.00Eþ00 0.00Eþ00 0.00Eþ00 0.00Eþ00 1.11E+01 1.14E+01 8.22E-14 0.00Eþ00
Worst 6.98E+02 9.62E+02 1.87E+02 1.43E+03 5.74E+02 1.18E+03 1.16E+03 1.15E+03 0.00Eþ00 4.14E+02 1.42E+02 4.19E+01 4.64E+02 6.10E+02 2.21E+01 0.00Eþ00
Avg 9.99E+01 2.52E+02 1.88E+01 2.80E+02 9.53E+01 2.41E+02 1.60E+02 2.41E+02 0.00Eþ00 5.97E+01 1.42E+00 3.16E+00 1.56E+02 1.77E+02 3.33E+00 0.00Eþ00
Std 1.54E+02 1.91E+02 2.63E+01 2.86E+02 1.16E+02 2.39E+02 2.44E+02 3.06E+02 0.00Eþ00 1.12E+02 1.42E+01 8.74E+00 1.07E+02 1.28E+02 5.87E+00 0.00Eþ00
CF9 Best 8.88E-15 5.91E-05 8.88E-15 0.00Eþ00 3.25E-05 1.41E-03 2.40E-08 1.50E-01 8.34E-33 2.01E-11 6.16E-12 1.45E-127 4.94E-05 4.65E-05 2.73E-08 2.41E-296
Worst 4.84E+00 6.70E+01 1.05E+01 7.88E+01 7.41E+01 8.58E+00 4.73E+00 8.17E+00 8.88E-15 9.91E+02 9.71E-08 3.51E-114 5.12E+01 6.76E+01 7.76E-06 5.80E-230
Avg 1.42E-01 5.56E+00 3.26E+00 1.51E+01 3.54E+00 1.36E+00 1.37E-01 1.14E+00 8.79E-15 1.52E+02 5.58E-09 4.85E-116 2.02E+00 2.30E+00 3.65E-07 7.31E-232
Std 8.12E-01 1.54E+01 3.44E+00 1.67E+01 1.02E+01 2.42E+00 7.86E-01 1.88E+00 8.88E-16 3.54E+02 1.27E-08 3.61E-115 5.52E+00 8.54E+00 1.01E-06 0.00Eþ00
F10 Best 4.81E+01 4.83E+01 4.83E+01 0.00Eþ00 1.15E+01 4.83E+01 4.82E+01 4.82E+01 9.18E-04 6.15E-02 1.80E-03 9.81E-05 2.17E-02 7.35E-03 4.85E+01 2.77E-277
Worst 8.21E+01 8.88E+01 5.11E+01 1.81E+02 4.98E+01 8.10E+01 8.24E+01 8.07E+01 5.09E+01 8.64E+01 8.65E+01 5.57E+01 1.07E+02 9.15E+01 4.95E+01 3.64E-212
Avg 5.07E+01 5.03E+01 4.92E+01 6.62E+01 4.81E+01 5.04E+01 4.98E+01 4.93E+01 1.40E+01 5.50E+01 4.94E+01 9.43E+00 5.25E+01 5.27E+01 4.91E+01 3.97E-214

Expert Systems With Applications 239 (2024) 122413

Std 7.47E+00 6.20E+00 5.90E-01 3.19E+01 4.51E+00 5.55E+00 5.41E+00 3.20E+00 2.07E+01 1.52E+01 2.59E+01 1.62E+01 1.18E+01 1.21E+01 2.08E-01 0.00Eþ00
CF11 Best 100.00 100.01 100.02 100.00 100.00 100.70 100.00 1136.75 100.00 100.00 100.00 100.00 100.13 100.03 100.00 100.00
Worst 100.00 6831.36 3960.34 72786433.70 100.00 10086.80 10100.00 16911.14 100.00 100.00 100.00 100.00 4673.64 5264.36 100.15 100.00
Avg 100.00 1256.91 562.85 3957605.24 100.00 2047.71 4000.02 7357.93 100.00 100.00 100.00 100.00 735.18 683.43 100.00 100.00
Std 1.03E-14 1.31E+03 6.70E+02 1.07E+07 0.00E+00 2.81E+03 4.90E+03 3.92E+03 0.00Eþ00 3.12E-07 3.57E-11 0.00E+00 9.22E+02 7.90E+02 1.51E-02 0.00Eþ00
CF12 Best 1100.25 1106.89 1100.13 1100.00 1733.88 1100.25 1107.02 1103.92 1100.00 1100.06 1100.00 1100.00 1100.06 1100.06 1100.06 1100.00
Worst 2141.91 2505.36 1688.04 2866.98 2629.67 2977.58 2407.48 2273.42 1100.53 1116.95 2620.05 1146.43 2304.67 2146.46 1543.34 1100.00
Avg 1390.90 1820.15 1245.02 1595.42 2226.17 1949.77 1561.16 1608.21 1100.01 1105.32 1126.42 1104.79 1546.63 1520.08 1147.44 1100.00
Std 2.16E+02 3.33E+02 1.68E+02 4.07E+02 1.81E+02 3.58E+02 3.09E+02 2.53E+02 5.64E-02 3.73E+00 1.59E+02 6.28E+00 2.99E+02 2.92E+02 7.36E+01 0.00Eþ00
CF13 Best 703.98 710.95 705.97 700.00 723.58 704.98 708.96 708.03 700.00 700.00 700.00 700.00 704.98 704.98 700.88 700.00
Worst 744.38 771.09 730.74 865.31 738.50 764.67 739.59 752.49 700.00 712.31 797.42 717.71 760.69 757.71 739.64 700.00
Avg 720.98 731.02 716.38 729.28 730.60 723.99 721.80 724.07 700.00 710.23 712.12 708.54 724.22 725.06 723.15 700.00
Std 7.98E+00 1.02E+01 4.71E+00 2.43E+01 3.64E+00 1.01E+01 6.58E+00 7.58E+00 0.00Eþ00 3.11E+00 2.91E+01 5.50E+00 1.23E+01 1.19E+01 7.31E+00 0.00Eþ00
CF14 Best 1900.40 1900.34 1900.40 1900.00 1900.89 1900.32 1900.47 1900.45 1900.00 1900.09 1900.00 1900.00 1900.17 1900.22 1900.98 1900.00
Worst 1902.32 1902.69 1902.44 1913.00 1902.93 1903.20 1903.12 1902.55 1900.00 1901.77 1904.80 1900.29 1903.57 1904.36 1902.42 1900.00
(continued on next page)
 M. Han et al.
Table 12 (continued )
Function AFT ALO CSA DA DMOA GOA MFO MVO NGO SA SCA SO SS SSA WSO WO

Avg 1901.03 1901.29 1900.99 1902.17 1902.16 1901.47 1901.18 1901.28 1900.00 1900.68 1900.39 1900.02 1901.28 1901.44 1901.90 1900.00
Std 3.58E-01 5.49E-01 4.08E-01 1.95E+00 4.11E-01 6.09E-01 5.64E-01 4.66E-01 2.28E-04 2.78E-01 1.03E+00 5.39E-02 6.07E-01 7.55E-01 3.26E-01 0.00Eþ00
CF15 Best 1701.41 2339.99 1702.78 1700.00 1705.02 2380.69 1700.00 1807.64 1700.00 1700.00 1700.00 1700.00 2091.75 1847.09 1700.00 1700.00
Worst 2259.24 15745.85 2134.63 83615.26 1744.88 14710.54 2540.16 3141.40 1700.00 1818.44 1700.00 1976.90 11155.53 11871.52 1711.38 1700.00
Avg 1816.93 4886.03 1801.11 5418.03 1716.47 7135.67 1782.95 2311.84 1700.00 1704.79 1700.00 1743.21 3786.89 3685.05 1701.60 1700.00
Std 1.26E+02 2.62E+03 1.12E+02 1.01E+04 8.13E+00 3.58E+03 1.34E+02 2.87E+02 0.00Eþ00 1.36E+01 6.09E-06 5.41E+01 1.72E+03 1.57E+03 1.78E+00 0.00Eþ00
CF16 Best 1600.33 1601.23 1600.16 1600.00 1601.29 1601.74 1600.39 1601.75 1600.00 1600.10 1600.00 1600.00 1601.01 1600.63 1600.37 1600.00
Worst 1730.71 1779.35 1616.49 1839.05 1605.00 1883.52 1756.02 1858.60 1600.31 1623.36 1615.82 1604.89 1738.26 1737.26 1611.85 1600.00
Avg 1611.80 1633.50 1602.97 1635.96 1602.69 1687.73 1623.18 1651.10 1600.02 1601.59 1600.39 1601.03 1610.85 1611.70 1602.73 1600.00
Std 2.17E+01 3.97E+01 2.53E+00 5.10E+01 6.97E-01 7.72E+01 3.05E+01 6.05E+01 3.53E-02 3.26E+00 1.66E+00 9.26E-01 1.66E+01 2.05E+01 2.48E+00 0.00Eþ00
CF17 Best 2100.08 2254.54 2100.62 2100.00 2100.97 2275.27 2100.12 2117.03 2100.00 2100.02 2100.00 2100.01 2119.46 2137.64 2100.03 2100.00
Worst 2366.15 13997.10 2407.73 15759.04 2103.31 15423.09 2538.88 2672.93 2100.00 2134.11 2100.26 2219.25 10750.27 10692.16 2102.17 2100.00
Avg 2155.18 4986.52 2128.32 3614.09 2101.71 6754.74 2124.27 2236.98 2100.00 2102.58 2100.03 2108.54 3254.18 3218.11 2100.42 2100.00
Std 6.93E+01 2.57E+03 4.91E+01 2.82E+03 5.03E-01 3.36E+03 4.93E+01 1.07E+02 2.32E-04 5.94E+00 5.32E-02 2.09E+01 1.29E+03 1.29E+03 3.33E-01 0.00Eþ00
24

CF18 Best 2200.00 2220.33 2200.00 2200.00 2200.02 2220.76 2200.00 2209.93 2200.00 2200.00 2200.00 2200.00 2219.85 2209.90 2200.00 2200.00
Worst 2926.85 3127.98 2310.10 3272.34 2936.43 2980.61 3370.39 3247.49 2200.00 2579.28 3777.11 2263.13 2901.68 2750.20 2223.58 2200.00
Avg 2308.99 2455.02 2221.41 2468.89 2326.90 2437.48 2387.30 2433.89 2200.00 2277.94 2215.77 2208.95 2362.73 2341.61 2202.17 2200.00
Std 1.60E+02 1.88E+02 2.26E+01 2.68E+02 1.43E+02 1.74E+02 2.51E+02 2.84E+02 0.00Eþ00 1.24E+02 1.58E+02 1.21E+01 1.16E+02 1.21E+02 4.72E+00 0.00Eþ00
CF19 Best 2400.00 2400.00 2400.00 2400.00 2400.00 2400.00 2400.00 2400.17 2400.00 2400.00 2400.00 2400.00 2400.00 2400.00 2400.00 2400.00
Worst 2405.62 2473.68 2409.03 2468.38 2473.01 2408.76 2405.61 2406.85 2400.00 3391.15 2400.00 2400.00 2411.74 2408.37 2400.00 2400.00
Avg 2400.11 2406.97 2403.33 2413.86 2406.21 2401.83 2400.06 2401.39 2400.00 2591.33 2400.00 2400.00 2401.51 2401.43 2400.00 2400.00
Std 7.71E-01 1.69E+01 3.20E+00 1.67E+01 1.20E+01 2.81E+00 5.61E-01 2.01E+00 3.60E-13 3.89E+02 1.33E-08 6.46E-14 2.66E+00 2.45E+00 2.15E-06 0.00Eþ00
CF20 Best 2548.06 2548.28 2548.27 2500.00 2509.55 2548.32 2548.17 2548.15 2500.00 2500.04 2500.00 2500.00 2535.02 2500.02 2548.26 2500.00
Worst 2581.35 2578.91 2557.91 2666.06 2549.77 2583.07 2579.28 2581.81 2550.89 2592.21 2586.88 2556.11 2608.87 2602.77 2549.59 2500.00
Avg 2550.01 2549.76 2549.27 2562.78 2547.55 2550.46 2549.75 2549.86 2513.14 2552.69 2551.83 2511.02 2552.65 2554.60 2549.07 2500.00
Std 6.17E+00 3.06E+00 1.11E+00 3.11E+01 5.98E+00 5.93E+00 5.10E+00 5.15E+00 2.09E+01 1.18E+01 2.57E+01 1.77E+01 1.10E+01 1.40E+01 2.25E-01 0.00Eþ00
(W|T|L) (0|20|0) (0|20|0) (0|20|0) (0|9|11) (0|20|0) (0|17|3) (0|20|0) (0|18|2) (6|14|0) (0|18|2) (0|18|2) (0|20|0) (0|20|0) (0|20|0) (0|20|0) (20|0|0)
Friedman mean rank 8.50 12.30 7.15 15.40 6.43 12.80 10.35 10.75 3.25 7.20 8.15 5.23 10.90 11.35 5.05 1.20

Expert Systems With Applications 239 (2024) 122413

Final rank 9 14 6 16 5 15 10 11 2 7 8 4 12 13 3 1
 Table 13
Results of the Wilcoxon rank-sum test on CEC 2021 benchmark suite (CF1-CF20).
M. Han et al.

Function AFT-WO ALO-WO CSA-WO DA-WO DMOA-WO GOA-WO MFO-WO MVO-WO NGO-WO SA-WO SCA-WO SO-WO SS-WO SSA-WO WSO-WO

P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H P-value H

CF1 5.64E-39 + 5.64E-39 + 5.64E-39 + 6.99E-33 + 5.64E-39 + 5.64E-39 + 3.48E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 +
CF2 5.62E-39 + 5.64E-39 + 5.59E-39 + 4.07E-36 + 5.64E-39 + 5.64E-39 + 5.63E-39 + 5.64E-39 + 1.58E-03 + 5.64E-39 + 8.28E-31 + 1.46E-35 + 5.62E-39 + 5.62E-39 + 5.64E-39 +
CF3 5.64E-39 + 5.64E-39 + 5.63E-39 + 3.04E-37 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + N/A = 5.62E-39 + 5.64E-39 + 5.64E-39 + 5.61E-39 + 5.62E-39 + 5.64E-39 +
CF4 5.64E-39 + 5.64E-39 + 5.64E-39 + 3.04E-37 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 7.32E-03 + 5.64E-39 + 2.68E-12 + 2.04E-04 + 5.64E-39 + 5.64E-39 + 5.64E-39 +
CF5 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.64E-29 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.53E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 +
CF6 2.56E-34 + 2.56E-34 + 2.56E-34 + 1.61E-30 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 +
CF7 5.64E-39 + 5.64E-39 + 5.64E-39 + 3.04E-37 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 +
CF8 2.16E-38 + 5.64E-39 + 5.62E-39 + 2.16E-38 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + N/A = 1.82E-34 + 7.49E-23 + 2.72E-06 + 5.64E-39 + 5.64E-39 + 5.64E-39 +
CF9 3.48E-41 + 5.64E-39 + 5.28E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.76E-45 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 +
CF10 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.64E-29 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 + 2.56E-34 +
CF11 N/A = 5.64E-39 + 5.64E-39 + 4.80E-35 + N/A = 5.64E-39 + 1.80E-13 + 5.64E-39 + N/A = 4.70E-39 + N/A = N/A = 5.64E-39 + 5.64E-39 + 5.74E-26 +
CF12 5.63E-39 + 5.64E-39 + 5.58E-39 + 3.03E-37 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 8.28E-03 + 5.64E-39 + 9.14E-10 + 8.08E-32 + 5.62E-39 + 5.60E-39 + 5.64E-39 +
CF13 5.64E-39 + 5.64E-39 + 5.62E-39 + 1.12E-36 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + N/A = 5.64E-39 + 3.68E-09 + 4.07E-36 + 5.61E-39 + 5.60E-39 + 5.64E-39 +
CF14 5.64E-39 + 5.64E-39 + 5.64E-39 + 4.07E-36 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 1.58E-03 + 5.64E-39 + 5.53E-08 + 5.10E-06 + 5.64E-39 + 5.64E-39 + 5.64E-39 +
CF15 5.63E-39 + 5.64E-39 + 5.64E-39 + 3.04E-37 + 5.64E-39 + 5.64E-39 + 2.15E-38 + 5.64E-39 + N/A = 5.63E-39 + 5.53E-08 + 2.16E-38 + 5.64E-39 + 5.64E-39 + 8.14E-38 +
CF16 5.64E-39 + 5.64E-39 + 5.64E-39 + 4.07E-36 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 +
CF17 5.64E-39 + 5.64E-39 + 5.64E-39 + 3.04E-37 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.63E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 +
CF18 1.12E-36 + 5.64E-39 + 8.30E-29 + 4.07E-36 + 5.64E-39 + 5.64E-39 + 2.16E-38 + 5.64E-39 + 5.63E-39 + 5.62E-39 + 5.64E-39 + 7.54E-23 + 5.64E-39 + 5.64E-39 + 5.82E-13 +
CF19 1.12E-36 + 5.62E-39 + 5.60E-18 + 2.16E-33 + 5.64E-39 + 5.64E-39 + 5.07E-08 + 5.64E-39 + 5.63E-39 + 5.18E-39 + 5.64E-39 + 7.54E-23 + 5.59E-39 + 5.60E-39 + 1.23E-03 +
CF20 5.64E-39 + 5.64E-39 + 5.64E-39 + 4.07E-36 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 + 5.64E-39 +

25
Subject to
Subject to

as follows:

⎩
⎪
⎪
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎧
⎩
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎧
respectively).

⎨ g2 =
Subject to g(x) =

g4 =

3
4
Minimize f (x) = x21 x2 (2 + x3 )

g3 = 1 −
g1 = 1 −

4x22 − x1 x2

1.5
12566(x2 x31 − x41

x1 + x2
Suppose x = [x1 x2 x3 ] = [dDN]

4.4.5. Case 5: Welded beam design

x22 x3
x32 x3

)+

4.4.4. Case 4: Pressure vessel design

71785x41

140.45x1
60 27 19 7 1

g4 = x4 − 240 ≤ 0
x31 x23 x33 x34 x35

Suppose x = [x1 x2 x3 x4 ] = [TsThRL]

g1 = 0.0193x3 − x1 ≤ 0
g2 = 0.00954x3 − x3 ≤ 0
− 1≤0
≤0

≤0
5108x21
+ + + + − 1≤0

g3 = − πx23 x4 − πx33 + 1296000 ≤ 0

4.4.3. Case 3: tension/compression spring design

WO is ranked as the first best solution obtained.

− 1≤0
where, x1,2,3,4,5 ∈ [0.01, 100] are design variables.

with a cost of 0.012665, which no other can achieve.

where,x1,2 ∈ [0.1, 99] and x3,4 ∈ [10, 200] are design variables.
Minimize f (x) = 0.6224x1 x3 x4 + 1.7781x2 x23 + 3.1661x21 x4 + 19.84x12 x3
where,x1 ∈ [0.05, 2], x2 ∈ [0.25, 1.3], x3 ∈ [2, 15] are design variables.
(g4 ). The mathematical model of spring design is described as follows:

The minimum cost is the objective function of the welded beam

The results obtained by WO and other metaheuristic algorithms

reported in Table 19. In this table, the comparison results show that the
WO and their comparison with the aforementioned metaheuristics are
cylinder length (L). The objective function is constrained by four
WO and HHO produced the best solution and achieved the best design
(g1 ), shear stress (g2 ), impact frequency (g3 ), and outer diameter limit
the spring (see Fig. 11(c)). The constraints include minimum deviation
coil diameter (d) and effective coil number (N) to minimize the weight of
Approximation (GCA) in its version I and II (GCA I and GCA II,
CS, Method of Moving Asymptotes (MMA), Generalized Convex
ranking order is ALO, MVO, Symbiotic Organisms Search (SOS), MFO,
SMA has the similar performance (i.e., 13.339957). The remaining

(HS) and Gravitational Search Algorithm (GSA). The results obtained by

Harris Hawks Optimization (HHO), ACO, WOA, GA, MVO, Co-
Some of the algorithms that are chosen for comparison are SMA,
constraint functions. The mathematical model of welding beam design is
optimized: shell thickness (Ts), head thickness (Th), inner radius (R) and
can be seen that the optimal cost obtained by WO (i.e., 13.013640) and
collected from the literatures are given in Table 17. From this table, it

evolutionary Particle Swarm Optimization (CPSO), Harmony Search

(30)
(29)
drical vessel design (see Fig. 11(d)). There are four parameters to be
Examining the optimal costs in Table 18, we can clearly observe that

The purpose of this design is to find the minimum cost of the cylin­
(28)
(27)
This work was done by optimizing the wire diameter (D), average
(26)
Expert Systems With Applications 239 (2024) 122413
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Fig. 9. Iteration diagram of CEC 2021 benchmark suite results.

[ ] [ ]
design problem (see Fig. 11(e)), and the objective function is con­ ( )√̅̅̅̅̅̅̅
x23 x64
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
strained by seven inequalities. The decision variables are weld thickness tion, Pc (x)= { 4.013 30 × 106 36 /196 × 1 − [x3
30×106
4(12×106 )
]/28
(h), steel bar connection length (l), steel bar height (t) and steel bar
is critical buckling load.
thickness (b) respectively. The mathematical model of welding beam
The WO was tested on this problem and the results were compared
design is as follows:
with AO, HPO, MPA, HHO, an effective Co-evolutionary Differential
Suppose x = [x1 x2 x3 x4 ] = [hltb]
Evolution (CDE), GSA, SIMPLEX, APPROX and DAVID. The results are
Minimizef (x) = 1.10471x21 x2 + 0.04811x3 x4 (14 + x2 )# (31) shown in Table 20. The WO has discovered better optimal values than
other algorithms and has been able to rank first.
⎧
⎪
⎪
⎪
⎪
⎪
⎪ g1 = τ(X) − 13600 ≤ 0 4.4.6. Case 6: Speed reducer design
⎪
⎪
⎪
⎪
⎪ g 2 = σ (X) − 30000 ≤ 0 The problem of minimizing the weight of a speed reducer involves
⎪
⎪
⎪
⎪ g3 = x1 − x4 ≤ 0 seven design variables and four constraints (see Fig. 11(f)). The design
⎨
Subject to g4 = 0.10471x21 + 0.04811x3 x4 (14 + x2 ) − 5 ≤ 0 (32) variables include the face width (x1 ), the module of teeth (x2 ), a discrete
⎪
⎪
⎪ design variable on behalf of the teeth in the pinion (x3 ), length and
⎪ g5 = 0.125 − x1 ≤ 0
⎪
⎪
⎪
⎪ diameter of the first shaft between bearings (x4 and x6 ), length and
⎪ g 6 = δ(X) − 0.25 ≤ 0
⎪
⎪
⎪ diameter of the second shaft between bearings (x5 and x7 ). Four con­
⎪
⎪
⎪
⎩
g 7 = 6000 − Pc (x) ≤ 0 straints covering stress (g1 ), bending stress of the gear teeth (g2 ), stresses
in the shafts (g3 ), and transverse deflections of the shafts (g4 ). The
mathematical model of welding beam design is as follows:
where,x1,4 ∈ [0.1, 2] and x2,3 ∈ [0.1, 10] are design variables, τ(X) =
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ Suppose x = [x1 x2 x3 x4 x5 x6 x7 ]
(τ′)2 + 2τ′τ″2R x2
+ (τ″)2 is shear stress, τ′ = √6000 ̅̅ , τ″ = MR
2x1 x2 J , M = ( )
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ Minimize f (x) =0.7854x1 x22 3.3333x23 + 14.9334x3 − 43.0934
( )
x22
( ) 2 √̅̅̅ x2 ( 2 ) ( ) (33)
x1 +x3 3 2
6000 14 +x22 , R = 4 + 2 , J = 2{ 2x1 x2 [122 + (x1 +x
2 ) ]},
− 1.508x1 x6 + x27 + 7.4777 x36 + x37
σ (X) = 504000
x2 x4
is beam bending stress, δ(X) = 65856000
is beam deflec­
3 (30×106 )x33 x4

26
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Fig. 10. Boxplot of test results of CEC 2021 benchmark suite.

Table 14
The original data of the simulation example.
Parameters Value

P = {p1 , p2 , ⋯, pM } {72,490,651,833,833,489,359,337,267,441,70,934,467,661,220,329,440,774,595,98,424,37,807,320,501,309,834,851,34,459,111,253,159,858,793,145,
651, 856, 400, 285, 405, 95, 391, 19, 96, 273, 152, 473, 448, 231}
W = {w1 , w2 , ⋯, {438,754,699,587,789,912,819,347,511,287,541,784,676,198,572,914,988,4,355,569,144,272,531,556,741,489,321,84,194,483,205,607,399,747,118,
wM } 651, 806, 9, 607, 121, 370,999,494,743,967,718,397,589,193,369}
V 11258

27
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Table 15
The original data of the simulation example.
Algorithms Solution Function value

GA (Jin & Ma, 2004) {10111000000111100110101010111110111011010000000111} 14865

SA (Jin & Ma, 2004) {01111101010101000110101010110110011011011000000110} 15844
BPSO (Ma et al., 2006) {01111001010111000110101010110100011011011010000110} 16052
WO {01111001010111000110101010110100011011011010000110} 16052

Fig. 11. Description of engineering design problems.

⎧
⎪
⎪
⎪
⎪
⎪
⎪ 27 397.5
⎪
⎪ g1 = − 1 ≤ 0, g2 = − 1≤0
⎪
⎪ x x 2
x x 2 2
1 x2 x3
⎪
⎪
1 2 3
⎪
⎪
⎪
⎪ 1.93x34 1.93x35
⎪
⎪ g3 = − 1 ≤ 0, g4 = − 1≤0
⎪
⎪ 4
⎪
⎪ x x
2 3 6 x x2 x3 x47
⎪
⎪ √(
̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
̅ √( ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
̅
⎪
⎪ )2 )2
⎪
⎨ 745x4 745x4
(34)
6 6
Subjectto + 16.9 × 10 + 16.9 × 10 #
⎪ x2 x3 x2 x3
⎪
⎪ g5 =
⎪
⎪ 110.0x6 3
− 1 ≤ 0, g6 =
110.0x6 3
− 1≤0
⎪
⎪
⎪
⎪
⎪
⎪ x
⎪ 2 x3 5x 2 x1
⎪
⎪ g7 = − 1 ≤ 0, g8 = − 1 ≤ 0, g9 = − 1≤0
⎪
⎪ 40 x1 12x2
⎪
⎪
⎪
⎪
⎪
⎪ 1.5x6 + 1.9 1.1x7 + 1.9
⎪
⎪ g10 = − 1 ≤ 0, g11 = − 1≤0
⎪
⎪ x4 x5
⎪
⎩

weight and shows the high performance of the WO is demonstrated by

its ability to approximate the global optimum for this problem.
where x1 ∈ [2.6, 3.6], x2 ∈ [0.7, 0.8], x3 ∈ [17, 28], x4 ∈ [7.3, 8.3], x5 ∈ In addition, the best, mean, worst and standard deviation of WO after
[7.8, 8.3], x6 ∈ [2.9, 3.9], and x7 ∈ [5.0, 5.5] are design variables. 10 independent runs were obtained. The statistical results of WO are
This problem has been solved by WO and compared with the liter­ shown in Table 22.
ature in Table 21. The comparison is made between Method of GWO,
AO, multidisciplinary design optimization (MDO), HS, hybrid Harris 5. Conclusions
Hawks-Sine Cosine Algorithm (hHHO-SCA), SCA, GSA, GA and PSO.
Note that WO outperforms other techniques when obtaining the lowest In this article, we present a nature-inspired migration and

28
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Table 16
Comparison of optimization results of gear train design
Algorithms Optimal values for variables f(x)

GA GB GC GD

WO 43 16 19 43 2.700857E-12
CS (Gandomi et al., 2013) 43 16 19 49 2.700900E-12
CSA (Gandomi et al., 2013) 19 16 43 49 2.701000E-12
WOA (Mirjalili & Lewis, 2016) 47 12 13 23 9.921570E− 10
GeneAS (Deb & Goyal, 1996) 33 14 17 50 1.362000E-09
Simulated annealing (Zhang & Wang, 1993) 52 15 30 60 2.360000E-09
Mixed integer discrete continuous optimization (Kannan & Kramer, 1994) 33 15 13 41 2.146000E-08
Augmented Lagrange Multiplier (Kannan & Kramer, 1994) 13 15 33 41 2.146000E-08
Mixed integer discrete continuous programming (Fu et al., 1991) 47 29 14 59 4.500000E-06
Nonlinear integer and discrete programming (Sandgren, 1990) 18 22 45 60 5.712000E-06

Table 17
Comparison of optimization results of cantilever beam design
Algorithms Optimal values for variables f(x)

x1 x2 x3 x4 x5

WO 5.947300 4.866000 4.434143 3.474900 2.186440 13.013640

SMA (Li et al., 2020) 6.017757 5.310892 4.493758 3.501106 2.150159 13.339957
ALO (Naruei et al., 2022) 6.018120 5.311420 4.488360 3.497510 2.158329 13.399500
MVO (Naruei et al., 2022) 6.023940 5.306011 4.495011 3.496022 2.152726 13.399595
SOS (Naruei et al., 2022) 6.018780 5.303440 4.495870 3.498960 2.155640 13.399600
MFO (Mirjalili, 2015a) 5.983000 5.316700 4.497300 3.513600 2.161600 13.399880
CS (Naruei et al., 2022) 6.008900 5.304900 4.502300 3.507700 2.150400 13.399900
MMA (Peraza-Vázquez et al., 2021) 6.010000 5.300000 4.490000 3.490000 2.150000 13.400000
GCA I (Peraza-Vázquez et al., 2021) 6.010000 5.300000 4.490000 3.490000 2.150000 13.400000
GCA II (Peraza-Vázquez et al., 2021) 6.010000 5.304000 4.490000 3.498000 2.150000 13.400000

reproduction of the famous animal - walrus. This mammal has adapted proposed algorithm, we have modeled global search simulating migra­
to land and sea environments with social intelligence. We propose a tion behavior of walruses and local search simulating reproduction
model-independent novel metaheuristic algorithm models behavior of behavior of walruses. Moreover, we have also modeled landing process
walrus into optimization technique and called Walrus Optimizer. In the and feeding process during the reproduction.

Table 18
Comparison of optimization results of tension/compression spring design.
Algorithms Optimal values for variables f(x)

d D N

WO 0.050000 0.311500 14.892300 0.012665

HHO (Heidari et al., 2019) 0.051796 0.359305 11.138859 0.012665
SHO (Naruei et al., 2022) 0.051144 0.343751 12.09550 0.012674
WOA (Naruei et al., 2022) 0.051207 0.345215 12.004032 0.012676
RO (Abualigah et al., 2022) 0.051370 0.349096 11.762790 0.012679
ES (Abualigah et al., 2022) 0.051643 0.355360 11.397926 0.012698
GSA (Naruei et al., 2022) 0.050276 0.323680 13.525410 0.012702
GA (Coello, 2000) 0.051480 0.351661 11.632201 0.012705
MVO (Abualigah et al., 2022) 0.052510 0.376020 10.335130 0.012790
CC (J. S. Arora, 2004) 0.050000 0.315900 14.250000 0.012833

Table 19
Comparison of optimization results of pressure vessel design.
Algorithms Optimal values for variables f(x)

Ts Th R L

WO 0.778190 0.384659 40.320119 199.993131 5885.349088

SMA (Li et al., 2020) 0.793100 0.393200 40.671100 196.217800 5994.185700
HHO (Heidari et al., 2019) 0.817584 0.407293 42.091746 176.719635 6000.462590
ACO (Kaveh & Talatahari, 2010) 0.812500 0.437500 42.098353 176.637751 6059.725800
WOA (Naruei et al., 2022) 0.812500 0.437500 42.098270 176.638998 6059.741000
GA (Abualigah, et al., 2022) 0.812500 0.437500 42.097398 176.654050 6059.946340
MVO (Abualigah, et al., 2022) 0.812500 0.437500 42.090738 176.738690 6060.806600
CPSO (Abualigah, et al., 2022) 0.812500 0.437500 42.091266 176.746500 6061.077700
HS (Abualigah, et al., 2022) 1.125000 0.625000 58.290150 43.692680 7197.730000
GSA (Rashedi et al., 2009) 1.125000 0.625000 55.988660 84.454203 8538.835900

29
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Table 20
Comparison of optimization results of welded beam design.
Algorithms Optimal values for variables f(x)

h l t b

WO 0.168066 4.065890 9.997888 0.168071 1.587354

AO (Abualigah et al., 2021) 0.163100 3.365200 9.020200 0.206700 1.656600
HPO (Naruei et al., 2022) 0.198812 3.337754 9.192016 0.198833 1.670240
MPA(Faramarzi et al., 2020) 0.205728 3.470509 9.036624 0.205730 1.724853
HHO (Heidari et al., 2019) 0.204039 3.531061 9.027463 0.206147 1.731991
CDE (Naruei et al., 2022) 0.203137 3.542998 9.033498 0.206179 1.733462
GSA (Naruei et al., 2022) 0.182129 3.856979 10.00000 0.202376 1.879950
SIMPLEX (Abualigah et al., 2022) 0.279200 5.625600 7.751200 0.279600 2.530700
APPROX (Abualigah et al., 2022) 0.244400 6.218900 8.291500 0.244400 2.381500
DAVID (Abualigah et al., 2022) 0.243400 6.255200 8.291500 0.244400 2.384100

Table 21
Comparison of optimization results of speed reducer design
Algorithms Optimal values for variables f(x)

x1 x2 x3 x4 x5 x6 x7

WO 3.500 0.700 17.000 7.300 7.715 3.350 5.287 2994.471066

GWO (M. Braik et al., 2021) 3.507 0.700 17.000 7.381 7.816 3.358 5.287 3001.288000
AO (Abualigah et al., 2021) 3.502 0.700 17.000 7.310 7.748 3.364 5.299 3007.732800
MDO (Lu & Kim, 2010) 3.500 0.700 17.000 7.300 7.670 3.542 5.246 3019.583365
HS (Geem et al., 2001) 3.520 0.700 17.000 8.370 7.800 3.367 5.289 3029.002000
hHHO-SCA (Abualigah et al., 2022) 3.506 0.700 17.000 7.300 7.991 3.453 5.287 3029.873076
CA (M. Braik et al., 2021) 3.509 0.700 17.000 7.300 7.800 3.461 5.289 3030.563000
GSA (M. Braik et al., 2021) 3.600 0.700 17.000 8.300 7.800 3.370 5.289 3051.120000
GA (M. Braik et al., 2021) 3.510 0.700 17.000 8.350 7.800 3.362 5.288 3067.561000
PSO (Stephen et al., 2018) 3.500 0.700 17.000 7.518 7.783 3.351 5.287 3145.922000

Table 22
Statistical results of WO in real-world engineering optimization problems.
Problems Best Mean Worst Std

Case 1 2.700857E-12 5.943261E-10 2.357641E-09 7.953806E-10

Case 2 13.012130 13.012978 13.013670 6.078283E-04
Case 3 0.012665 0.013363 0.014771 7.069479E-04
Case 4 5885.334775 6179.253361 6486.857909 2.268564E+02
Case 5 1.587154 1.588547 1.591209 1.243997E-03
Case 6 2994.471071 2994.471272 2994.472253 3.705633E-04

A series of mathematical problems (23 benchmark functions, CEC problems. WO can be combined with different fields, applied to real
2021 benchmark suite, 0–1 knapsack problem and real-world engi­ problems, and provide more possibilities for solving complex problems.
neering optimization problems) are used to evaluate the proposed WO The future research is directed towards solving NP-hard problems
and compared with other metaheuristics. The statistics from the and optimizing the design of complex engineering cases using WO.
benchmark tests show rapid convergence. WO locates most individuals Considering the specific requirements and characteristics of the prob­
around the best value with a small number of iterations, which gives WO lem, the algorithm strategy and parameters are adjusted to achieve the
a good advantage in improving the accuracy of the result. In addition, best performance. In addition, a multi-objective optimization version of
the Friedman test and Wilcoxon rank sum test quantitatively verified the WO can be developed and applied to different problems. To explore the
advantages of WO over the fifteen comparison algorithms. WO optimi­ practicality and innovation of the new intelligent optimization algo­
zation performance is superior to other algorithms in different engi­ rithm in designing engineering cases.
neering examples.
WO maintains a leading position in terms of optimality (i.e., the CRediT authorship contribution statement
capacity to reach optimal values), scalability (i.e., ability to deal with
large-scale problems) and computation time (especially for large-scale Muxuan Han: Conceptualization, Software, Visualization, Writing –
problems). However, its limitations are shown in competitiveness original draft. Zunfeng Du: Methodology, Validation, Formal analysis,
compared to approximation algorithms in the performances of modeling Supervision. Kum Fai Yuen: Supervision, Writing – review & editing.
(i.e., capacity to develop models that accurately represent the real-life Haitao Zhu: Validation, Formal analysis. Yancang Li: Methodology,
system) and uncertainty (i.e., ability to cope with non-deterministic Formal analysis. Qiuyu Yuan: Methodology, Software, Visualization,
scenarios). Therefore, it also opens up some promising research di­ Formal analysis.
rections and potentials. If WO is chosen to optimize a specific case in a
subsequent study, WO can be further improved according to the specific Declaration of Competing Interest
situation and characteristics of relevant cases, so as to ensure that the
algorithm has better optimization performance when solving specific The authors declare that they have no known competing financial

30
M. Han et al. Expert Systems With Applications 239 (2024) 122413

interests or personal relationships that could have appeared to influence Braik, M., Hammouri, A., Atwan, J., Al-Betar, M. A., & Awadallah, M. A. (2022). White
Shark Optimizer: A novel bio-inspired meta-heuristic algorithm for global
the work reported in this paper.
optimization problems. Knowledge-Based Systems, 243, Article 108457. https://doi.
org/10.1016/j.knosys.2022.108457
Data availability Braik, M., Ryalat, M. H., & Al-Zoubi, H. (2022). A novel meta-heuristic algorithm for
solving numerical optimization problems: Ali Baba and the forty thieves. Neural
Computing and Applications, 34(1), 409–455. https://doi.org/10.1007/s00521-021-
Data will be made available on request. 06392-x
Braik, M. S. (2021). Chameleon Swarm Algorithm: A bio-inspired optimizer for solving
Acknowledgments engineering design problems. Expert Systems with Applications, 174, Article 114685.
https://doi.org/10.1016/j.eswa.2021.114685
Braik, M., Sheta, A., & Al-Hiary, H. (2021). A novel meta-heuristic search algorithm for
This research is supported by the National Natural Science Founda­ solving optimization problems: Capuchin Search Algorithm. Neural Computing and
tion of China [grant number 51109158, and grant number U2106223], Applications, 33(7), 2515–2547. https://doi.org/10.1007/s00521-020-05145-6
Brown, A. R., Roberts, D. A., Susskind, L., Swingle, B., & Zhao, Y. (2016). Complexity,
the Science and Technology Development Plan Program of Tianjin action, and black holes. Physical Review D, 93(8), Article 086006. https://doi.org/
Municipal Transportation Commission [grant number 2022-48], the 10.1103/PhysRevD.93.086006
Tianjin Research Innovation Project for Postgraduate Students [grant Chica, M., Juan, A. A., Bayliss, C., Cordon, O., & Kelton, W. D. (2017). Why
Simheuristics? Benefits, Limitations, and Best Practices When Combining
number 2022BKY077], and the Tianjin University Graduate Education Metaheuristics with Simulation. SSRN Electronic Journal, 44(2), 311–334. https://
Special Found 2021 Projects [grant number C1-2021-004]. doi.org/10.2436/20.8080.02.104
Chopra, N., & Mohsin Ansari, M. (2022). Golden Jackal Optimization: A novel nature-
inspired optimizer for engineering applications. Expert Systems with Applications, 198,
References
Article 116924. https://doi.org/10.1016/j.eswa.2022.116924
Chouhan, V. K., Khan, S. H., & Hajiaghaei-Keshteli, M. (2021). Metaheuristic approaches
Abdel-Basset, M., Mohamed, R., Jameel, M., & Abouhawwash, M. (2023). Nutcracker to design and address multi-echelon sugarcane closed-loop supply chain network.
optimizer: A novel nature-inspired metaheuristic algorithm for global optimization Soft Computing, 25, 11377–11404. https://doi.org/10.1007/s00500-021-05943-7
and engineering design problems. Knowledge-Based Systems, 262, Article 110248. Chouhan, V. K., Khan, S. H., & Hajiaghaei-Keshteli, M. (2022). Hierarchical tri-level
https://doi.org/10.1016/j.knosys.2022.110248 optimization model for effective use of by-products in a sugarcane supply chain
Abdollahzadeh, B., Gharehchopogh, F. S., & Mirjalili, S. (2021). African Vultures network. Applied Soft Computing, 128, Article 109468. https://doi.org/10.1016/j.
Optimization Algorithm: A new nature-inspired metaheuristic algorithm for global asoc.2022.109468
optimization problems. Computers & Industrial Engineering, 158, Article 107408. Chouhan, V. K., Khan, S.h., & Hajiaghaei-Keshteli, M. (2021). Sustainable planning and
https://doi.org/10.1016/j.cie.2021.107408 decision-making model for sugarcane mills considering environmental issues.
Abu Doush, I., Awadallah, M. A., Al-Betar, M. A., Alomari, O. A., Makhadmeh, S. N., Journal of Environmental Management, 303, Article 114252. https://doi.org/10.1016/
Abasi, A. K., & Alyasseri, Z. A. A. (2023). Archive-based coronavirus herd immunity j.jenvman.2021.114252
algorithm for optimizing weights in neural networks. Neural Computing and Coello, C. (2000). Use of a self-adaptive penalty approach for engineering optimization
Applications, 35(21), 15923–15941. https://doi.org/10.1007/s00521-023-08577-y problems. Computers in Industry, 41(2), 113–127. https://doi.org/10.1016/S0166-
Abualigah, L., & Diabat, A. (2020). A comprehensive survey of the Grasshopper 3615(99)00046-9
optimization algorithm: Results, variants, and applications. Neural Computing and Coello Coello, C. A. (2002). Theoretical and numerical constraint-handling techniques
Applications, 32(19), 15533–15556. https://doi.org/10.1007/s00521-020-04789-8 used with evolutionary algorithms: A survey of the state of the art. Computer Methods
Abualigah, L., Diabat, A., Mirjalili, S., Elaziz, M. A., & Gandomi, A. H. (2021). The in Applied Mechanics and Engineering, 191(11–12), 1245–1287. https://doi.org/
Arithmetic Optimization Algorithm. Computer Methods in Applied Mechanics and 10.1016/S0045-7825(01)00323-1
Engineering, 376, Article 113609. https://doi.org/10.1016/j.cma.2020.113609 Deb, K. (1991). Optimal design of a welded beam via genetic algorithms. Aiaa Journal, 29
Abualigah, L., Elaziz, M. A., Sumari, P., Geem, Z. W., & Gandomi, A. H. (2022). Reptile (11), 2013–2015. https://doi.org/10.2514/3.10834
Search Algorithm (RSA): A nature-inspired meta-heuristic optimizer. Expert Systems Deb, K., & Goyal, M. (1996). A combined Genetic Adaptive Search (GeneAS) for engineering
with Applications, 191, Article 116158. https://doi.org/10.1016/j.eswa.2021.116158 design. http://core.ac.uk/display/22285732.
Abualigah, L., Elaziz, M. A., Sumari, P., Khasawneh, A. M., Alshinwan, M., Mirjalili, S., Dehghani, M., Hubalovsky, S., & Trojovsky, P. (2021). Northern Goshawk Optimization:
Shehab, M., Abuaddous, H. Y., & Gandomi, A. H. (2022). Black hole algorithm: A A new swarm-based algorithm for solving optimization problems. IEEE Access, 9,
comprehensive survey. Applied Intelligence, 52(10), 11892–11915. https://doi.org/ 162059–162080. https://doi.org/10.1109/ACCESS.2021.3133286
10.1007/s10489-021-02980-5 Dehghani, M., Montazeri, Z., Trojovská, E., & Trojovský, P. (2023). Coati Optimization
Abualigah, L., Yousri, D., Abd Elaziz, M., Ewees, A. A., Al-qaness, M. A. A., & Algorithm: A new bio-inspired metaheuristic algorithm for solving optimization
Gandomi, A. H. (2021). Aquila Optimizer: A novel meta-heuristic optimization problems. Knowledge-Based Systems, 259, Article 110011. https://doi.org/10.1016/j.
algorithm. Computers & Industrial Engineering, 157, Article 107250. https://doi.org/ knosys.2022.110011
10.1016/j.cie.2021.107250 Dehghani, M., & Trojovský, P. (2023). Osprey optimization algorithm: A new bio-
Agushaka, J. O., Ezugwu, A. E., & Abualigah, L. (2022). Dwarf Mongoose Optimization inspired metaheuristic algorithm for solving engineering optimization problems.
Algorithm. Computer Methods in Applied Mechanics and Engineering, 391, Article Frontiers in Mechanical Engineering, 8, 1126450. https://doi.org/10.3389/
114570. https://doi.org/10.1016/j.cma.2022.114570 fmech.2022.1126450
Ahmadianfar, I., Heidari, A. A., Gandomi, A. H., & Chu, X. (2021). RUN beyond the Dhiman, G., & Kumar, V. (2017). Spotted hyena optimizer: A novel bio-inspired based
metaphor: An efficient optimization algorithm based on Runge Kutta method. Expert metaheuristic technique for engineering applications. Advances in Engineering
Systems with Applications, 181, Article 115079. https://doi.org/10.1016/j. Software, 114, 48–70. https://doi.org/10.1016/j.advengsoft.2017.05.014
eswa.2021.115079 Dhiman, G., & Kumar, V. (2019). Seagull Optimization Algorithm: Theory and its
Akdağ, O. (2022). A Developed Honey Badger Optimization Algorithm for Tackling applications for large-scale industrial engineering problems. Knowledge-Based
Optimal Power Flow Problem. Electric Power Components and Systems, 50(6–7), Systems, 165, 169–196. https://doi.org/10.1016/j.knosys.2018.11.024
331–348. https://doi.org/10.1080/15325008.2022.2136295 Eltamaly, A. M., & Rabie, A. H. (2023). A Novel Musical Chairs Optimization Algorithm.
Al-Betar, M. A., Alyasseri, Z. A. A., Awadallah, M. A., & Abu Doush, I. (2021). Arabian Journal for Science and Engineering, 1–33. https://doi.org/10.1007/s13369-
Coronavirus herd immunity optimizer (CHIO). Neural Computing and Applications, 33 023-07610-5
(10), 5011–5042. https://doi.org/10.1007/s00521-020-05296-6 Emami, H. (2022a). Anti-coronavirus optimization algorithm. Soft Computing, 26(11),
Alkayem, N. F., Cao, M., Shen, L., Fu, R., & Šumarac, D. (2022). The combined social 4991–5023. https://doi.org/10.1007/s00500-022-06903-5
engineering particle swarm optimization for real-world engineering problems: A Emami, H. (2022b). Stock Exchange Trading Optimization Algorithm: A human-inspired
case study of model-based structural health monitoring. Applied Soft Computing, 123, method for global optimization. Journal of Supercomputing, 78(2), 2125–2174.
Article 108919. https://doi.org/10.1016/j.asoc.2022.108919 https://doi.org/10.1007/s11227-021-03943-w
Arora, J. S. (2004). Introduction to optimum design. Elsevier. Faramarzi, A., Heidarinejad, M., Mirjalili, S., & Gandomi, A. H. (2020). Marine Predators
Arora, S., & Singh, S. (2019). Butterfly Optimization Algorithm: A novel approach for Algorithm: A nature-inspired metaheuristic. Expert Systems with Applications, 152,
global optimization. Soft Computing, 23(3), 715–734. https://doi.org/10.1007/ Article 113377. https://doi.org/10.1016/j.eswa.2020.113377
s00500-018-3102-4 Faridmehr, I., Nehdi, M. L., Davoudkhani, I. F., & Poolad, A. (2023). Mountaineering
Awad, N. H., Ali, M. Z., & Suganthan, P. N. (2017). Ensemble sinusoidal differential Team-Based Optimization: A Novel Human-Based Metaheuristic Algorithm.
covariance matrix adaptation with Euclidean neighborhood for solving CEC2017 Mathematics, 11(5), Article 5. https://doi.org/10.3390/math11051273
benchmark problems. IEEE Congress on Evolutionary Computation (CEC), 2017, Fathollahi-Fard, A. M., Hajiaghaei-Keshteli, M., & Tavakkoli-Moghaddam, R. (2018). The
372–379. https://doi.org/10.1109/CEC.2017.7969336 Social Engineering Optimizer (SEO). Engineering Applications of Artificial Intelligence,
Azizi, M. (2021). Atomic Orbital Search: A novel metaheuristic algorithm. Applied 72, 267–293. https://doi.org/10.1016/j.engappai.2018.04.009
Mathematical Modelling, 93, 657–683. https://doi.org/10.1016/j.apm.2020.12.021 Fathollahi-Fard, A. M., Hajiaghaei-Keshteli, M., & Tavakkoli-Moghaddam, R. (2020). Red
Azizi, M., Aickelin, U., Khorshidi, A., Baghalzadeh, H., & Shishehgarkhaneh, M. (2023). deer algorithm (RDA): A new nature-inspired meta-heuristic. Soft Computing, 24(19),
Energy valley optimizer: A novel metaheuristic algorithm for global and engineering 14637–14665. https://doi.org/10.1007/s00500-020-04812-z
optimization. Scientific Reports, 13(1), Article 1. https://doi.org/10.1038/s41598- Fathollahi-Fard, A. M., Wong, K. Y., & Aljuaid, M. (2023). An efficient adaptive large
022-27344-y neighborhood search algorithm based on heuristics and reformulations for the

31
M. Han et al. Expert Systems With Applications 239 (2024) 122413

generalized quadratic assignment problem. Engineering Applications of Artificial Kumar, C., Magdalin Mary, D., & Gunasekar, T. (2022). MOCHIO: A novel Multi-
Intelligence, 126, Article 106802. https://doi.org/10.1016/j.engappai.2023.106802 Objective Coronavirus Herd Immunity Optimization algorithm for solving brushless
Feng, Z., Niu, W., & Liu, S. (2021). Cooperation Search Algorithm: A novel metaheuristic direct current wheel motor design optimization problem. Automatika, 63(1),
evolutionary intelligence algorithm for numerical optimization and engineering 149–170. https://doi.org/10.1080/00051144.2021.2014035
optimization problems. Applied Soft Computing, 98, Article 106734. https://doi.org/ Kuo, W., & Prasad, V. R. (2000). An annotated overview of system-reliability
10.1016/j.asoc.2020.106734 optimization. IEEE Transactions on Reliability, 49(2), 176–187. https://doi.org/
Fu, J. F., Fenton, R. G., & Cleghorn, W. L. (1991). A mixed integer-discrete-continuous 10.1109/24.877336
programming method and its application to engineering design optimization. Ladyman, J., Lambert, J., & Wiesner, K. (2013). What is a complex system? European
Engineering Optimization, 17(4), 263–280. https://doi.org/10.1080/ Journal for Philosophy of Science, 3(1), 33–67. https://doi.org/10.1007/s13194-012-
03052159108941075 0056-8
Gandomi, A. H. (2014). Interior search algorithm (ISA): A novel approach for global Lakshmi, G. V. N., Jayalaxmi, A., & Veeramsetty, V. (2023). Optimal placement of
optimization. ISA Transactions, 53(4), 1168–1183. https://doi.org/10.1016/j. distributed generation based on DISCO’s financial benefit with loss and emission
isatra.2014.03.018 reduction using hybrid Jaya-Red Deer optimizer. Electrical Engineering, 105(2),
Gandomi, A. H., Yang, X.-S., & Alavi, A. H. (2013). Cuckoo Search Algorithm: A 965–977. https://doi.org/10.1007/s00202-022-01709-y
metaheuristic approach to solve structural optimization problems. Engineering with Li, S., Chen, H., Wang, M., Heidari, A. A., & Mirjalili, S. (2020). Slime Mould Algorithm:
Computers, 29(1), 17–35. https://doi.org/10.1007/s00366-011-0241-y A new method for stochastic optimization. Future Generation Computer Systems, 111
Geem, Z. W., Kim, J. H., & Loganathan, G. V. (2001). A new heuristic optimization aliasgharheidari.com, 300–323. https://doi.org/10.1016/j.future.2020.03.055.
algorithm: Harmony Search. Simulation, 76(2), 60–68. https://doi.org/10.1177/ Liang, J. J., Qu, B. Y., & Suganthan, P. N. (2014). Problem definitions and evaluation
003754970107600201 criteria for the CEC 2014 special session and competition on single objective real-
Gotfredsen, A. B., Appelt, M., & Hastrup, K. (2018). Walrus history around the North parameter numerical optimization [Technical Report].
Water: Human-animal relations in a long-term perspective. Ambio: A Journal of the Lim, T. Y., Al-Betar, M. A., & Khader, A. T. (2016). Taming the 0/1 knapsack problem
Human. Environment, 47(2), 193–212. https://doi.org/10.1007/s13280-018-1027-x with monogamous pairs genetic algorithm. Expert Systems with Applications, 54,
Hajiaghaei-Keshteli, M., & Aminnayeri, M. (2013). A new optimization algorithm 241–250. https://doi.org/10.1016/j.eswa.2016.01.055
inspired by Keshtels’ feeding. Proceeding in IEEE Conference on Industrial Engineering Lu, S., & Kim, H. M. (2010). A regularized inexact penalty decomposition algorithm for
and Management Systems, 2249–2253. multidisciplinary design optimization problems with complementarity constraints.
Hajiaghaei-Keshteli, M., & Aminnayeri, M. (2014). Solving the integrated scheduling of Journal of Mechanical Design, 132(4), Article 041005. https://doi.org/10.1115/
production and rail transportation problem by Keshtel algorithm. Applied Soft 1.4001206
Computing, 25, 184–203. https://doi.org/10.1016/j.asoc.2014.09.034 Ma, H., Ye, C., & Zhang, S. (2006). Binary improved particle swarm optimization
Han, M., Du, Z., Zhu, H., Li, Y., Yuan, Q., & Zhu, H. (2022). Golden-Sine dynamic marine algorithm for knapsack problem. Journal of University of Shanghai for Science and
predator algorithm for addressing engineering design optimization. Expert Systems Technology, 1, 31–34. https://doi.org/10.13255/j.cnki.jusst.2006.01.008.
with Applications, 210, 118460. https://doi.org/10.1016/j.eswa.2022.118460 Mirjalili, S. (2015a). Moth-Flame Optimization Algorithm: A novel nature-inspired
Harifi, S., Mohammadzadeh, J., Khalilian, M., & Ebrahimnejad, S. (2021). Giza Pyramids heuristic paradigm. Knowledge-Based Systems, 89, 228–249. https://doi.org/
Construction: An ancient-inspired metaheuristic algorithm for optimization. 10.1016/j.knosys.2015.07.006
Evolutionary Intelligence, 14(4), 1743–1761. https://doi.org/10.1007/s12065-020- Mirjalili, S. (2015b). The Ant Lion Optimizer. Advances in Engineering Software, 83,
00451-3 80–98. https://doi.org/10.1016/j.advengsoft.2015.01.010
Hashemi-Amiri, O., Mohammadi, M., Rahmanifar, G., Hajiaghaei-Keshteli, M., Fusco, G., Mirjalili, S. (2016a). Dragonfly Algorithm: A new meta-heuristic optimization technique
& Colombaroni, C. (2023). An allocation-routing optimization model for integrated for solving single-objective, discrete, and multi-objective problems. Neural
solid waste management. Expert Systems with Applications, 227, Article 120364. Computing and Applications, 27(4), 1053–1073. https://doi.org/10.1007/s00521-
https://doi.org/10.1016/j.eswa.2023.120364 015-1920-1
Hashim, F. A., Houssein, E. H., Hussain, K., Mabrouk, M. S., & Al-Atabany, W. (2022). Mirjalili, S. (2016b). SCA: A Sine Cosine Algorithm for solving optimization problems.
Honey badger algorithm: New metaheuristic algorithm for solving optimization Knowledge-Based Systems, 96, 120–133. https://doi.org/10.1016/j.
problems. Mathematics and Computers in Simulation, 192, 84–110. https://doi.org/ knosys.2015.12.022
10.1016/j.matcom.2021.08.013 Mirjalili, S., Gandomi, A. H., Mirjalili, S. Z., Saremi, S., Faris, H., & Mirjalili, S. M. (2017).
Hashim, F. A., & Hussien, A. G. (2022). Snake Optimizer: A novel meta-heuristic Salp Swarm Algorithm: A bio-inspired optimizer for engineering design problems.
optimization algorithm. Knowledge-Based Systems, 242, Article 108320. https://doi. Advances in Engineering Software, 114, 163–191. https://doi.org/10.1016/j.
org/10.1016/j.knosys.2022.108320 advengsoft.2017.07.002
Hayyolalam, V., & Pourhaji Kazem, A. A. (2020). Black Widow Optimization Algorithm: Mirjalili, S., & Lewis, A. (2016). The Whale Optimization Algorithm. Advances in
A novel meta-heuristic approach for solving engineering optimization problems. Engineering Software, 95, 51–67. https://doi.org/10.1016/j.advengsoft.2016.01.008
Engineering Applications of Artificial Intelligence, 87, Article 103249. https://doi.org/ Mirjalili, S., Mirjalili, S. M., & Hatamlou, A. (2015). Multi-Verse Optimizer: A nature-
10.1016/j.engappai.2019.103249 inspired algorithm for global optimization. Neural Computing and Applications, 27(2),
Heidari, A. A., Mirjalili, S., Faris, H., Aljarah, I., Mafarja, M., & Chen, H. (2019). Harris 495–513. https://doi.org/10.1007/s00521-015-1870-7
hawks optimization: Algorithm and applications. Future Generation Computer Systems, Mirjalili, S., Mirjalili, S. M., & Lewis, A. (2014). Grey Wolf Optimizer. Advances in
97, 849–872. https://doi.org/10.1016/j.future.2019.02.028 Engineering Software, 69, 46–61. https://doi.org/10.1016/j.advengsoft.2013.12.007
Ho, Y. C., & Pepyne, D. L. (2002). Simple explanation of the No-Free-Lunch theorem and Moghadam, P., & Ahmadi, A. (2023). A novel two-stage bio-inspired method using red
its implications. Journal of Optimization Theory and Applications, 115(3), 549–570. deer algorithm for data clustering. Evolutionary Intelligence, 1–18. https://doi.org/
https://doi.org/10.1023/A:1021251113462 10.1007/s12065-023-00864-w
Holland, J. H. (1992). Adaptation in natural and artificial systems: An introductory Mohamed, A. W., Sallam, K. M., Agrawal, P., Hadi, A. A., & Mohamed, A. K. (2023).
analysis with applications to biology, control, and artificial intelligence. MIT press. Evaluating the performance of meta-heuristic algorithms on CEC 2021 benchmark
http://ui.adsabs.harvard.edu/abs/1975anas.book....H. problems. Neural Computing and Applications, 35(2), 1493–1517. https://doi.org/
Hosseini, E., Sadiq, A. S., Ghafoor, K. Z., Rawat, D. B., Saif, M., & Yang, X. (2021). 10.1007/s00521-022-07788-z
Volcano Eruption Algorithm for solving optimization problems. Neural Computing Mohammadi-Balani, A., Dehghan Nayeri, M., Azar, A., & Taghizadeh-Yazdi, M. (2021).
and Applications, 33(7), 2321–2337. https://doi.org/10.1007/s00521-020-05124-x Golden Eagle Optimizer: A nature-inspired metaheuristic algorithm. Computers &
Jay, C. V., Olson, T. L., Garner, G. W., & Ballachey, B. E. (1998). Response of Pacific Industrial Engineering, 152, Article 107050. https://doi.org/10.1016/j.
walruses to disturbances from capture and handling activities at a haul-out in Bristol cie.2020.107050
Bay, Alaska. Marine Mammal Science, 14(4), 819–828. https://doi.org/10.1111/ Mosallanezhad, B., Chouhan, V.k., Paydar, M.m., & Hajiaghaei-Keshteli, M. (2021).
j.1748-7692.1998.tb00765.x Disaster relief supply chain design for personal protection equipment during the
Jin, H., & Ma, L. (2004). Genetic annealing evolutionary algorithm applied to the COVID-19 pandemic. Applied Soft Computing, 112, Article 107809. https://doi.org/
knapsack problem. Journal of University of Shanghai for Science and Technology, 6, 10.1016/j.asoc.2021.107809
561–564. https://doi.org/10.13255/j.cnki.jusst.2004.06.016. Mosallanezhad, B., Hajiaghaei-Keshteli, M., Smith Cornejo, N. R., & Rodríguez
Kannan, B. K., & Kramer, S. N. (1994). An augmented lagrange multiplier based method Calvo, E. Z. (2023). An IoMT platform for an integrated sustainable energy-efficient
for mixed integer discrete continuous optimization and its applications to disaster relief supply chain to prevent severity-driven disruptions during pandemics.
mechanical design. Journal of Mechanical Design, 116(2), 405–411. https://doi.org/ Journal of Industrial Information Integration, 35, Article 100502. https://doi.org/
10.1115/1.2919393 10.1016/j.jii.2023.100502
Karaboga, D., & Basturk, B. (2007). A powerful and efficient algorithm for numerical Mousavi, R., Salehi-Amiri, A., Zahedi, A., & Hajiaghaei-Keshteli, M. (2021). Designing a
function optimization: Artificial Bee Colony (ABC) algorithm. Journal of Global Supply Chain Network for Blood Decomposition by Utilizing Social and
Optimization, 39(3), 459–471. https://doi.org/10.1007/s10898-007-9149-x Environmental Factor. Computers & Industrial Engineering, 160(1), Article 107501.
Kaveh, A., & Talatahari, S. (2010). An improved ant colony optimization for constrained https://doi.org/10.1016/j.cie.2021.107501
engineering design problems. Engineering Computations, 27(1–2), 155–182. https:// Naruei, I., & Keynia, F. (2021a). A new optimization method based on COOT bird natural
doi.org/10.1108/02644401011008577 life model. Expert Systems with Applications, 183, Article 115352. https://doi.org/
Kennedy, J., & Eberhart, R. (1995). Particle swarm optimization. In Proceedings of 10.1016/j.eswa.2021.115352
ICNN’95 - International Conference on Neural Networks. http://ieeexplore.ieee.org/ Naruei, I., & Keynia, F. (2021b). Wild Horse Optimizer: A new meta-heuristic algorithm
xpl/articleDetails.jsp?reload=true&arnumber=488968. for solving engineering optimization problems. Engineering with Computers, 38,
Khishe, M., & Mosavi, M. R. (2020). Chimp Optimization Algorithm. Expert Systems with 3025–3056. https://doi.org/10.1007/s00366-021-01438-z
Applications, 149, Article 113338. https://doi.org/10.1016/j.eswa.2020.113338 Naruei, I., Keynia, F., & Sabbagh Molahosseini, A. (2022). Hunter-Prey Optimization:
Kirkpatrick, S., Gelatt, C. D., & Vecchi, A. (1983). Optimization by simulated annealing. Algorithm and applications. Soft Computing, 26(3), 1279–1314. https://doi.org/
Science, 220(4598), 671–680. https://doi.org/10.1126/science.220.4598.671 10.1007/s00500-021-06401-0

32
M. Han et al. Expert Systems With Applications 239 (2024) 122413

Olaide, O., Ezugwu, E. S., Mohamed, T., & Abualigah, L. (2022). Ebola Optimization Talatahari, S., Bayzidi, H., & Saraee, M. (2021). Social Network Search for global
Search Algorithm: A new nature-inspired metaheuristic optimization algorithm. IEEE optimization. IEEE Access, 9, 92815–92863. https://doi.org/10.1109/
Access, 10, 1–38. https://doi.org/10.1109/ACCESS.2022.3147821 ACCESS.2021.3091495
Ong, K. M., Ong, P., & Sia, C. K. (2021). A carnivorous plant algorithm for solving global Verij kazemi, M., & Fazeli Veysari, E. (2022). A new optimization algorithm inspired by
optimization problems. Applied Soft Computing, 98, Article 106833. https://doi.org/ the quest for the evolution of human society: Human felicity algorithm. Expert
10.1016/j.asoc.2020.106833 Systems with Applications, 193, Article 116468. https://doi.org/10.1016/j.
Pan, J.-S., Zhang, S., Chu, S., Yang, H., & Yan, B. (2023). Willow Catkin Optimization eswa.2021.116468
Algorithm Applied in the TDOA-FDOA Joint Location Problem. Entropy, 25(1), Wang, L., Cao, Q., Zhang, Z., Mirjalili, S., & Zhao, W. (2022). Artificial Rabbits
Article 1. https://doi.org/10.3390/e25010171 Optimization: A new bio-inspired meta-heuristic algorithm for solving engineering
Pan, W. T. (2012). A new Fruit Fly Optimization Algorithm: Taking the financial distress optimization problems. Engineering Applications of Artificial Intelligence, 114, Article
model as an example. Knowledge-Based Systems, 26, 69–74. https://doi.org/10.1016/ 105082. https://doi.org/10.1016/j.engappai.2022.105082
j.knosys.2011.07.001 Wang, T., Zhou, J., Zhang, Q., Lin, C., & Liang, Y. (2023). Design of Adaptive Time-
Peraza-Vázquez, H., Peña-Delgado, A. F., Echavarría-Castillo, G., Morales-Cepeda, A. B., Varying Sliding Mode Controller for Underactuated Overhead Crane Optimized via
Velasco-Álvarez, J., & Ruiz-Perez, F. (2021). A bio-inspired method for engineering Improved Honey Badger Algorithm. Journal of Intelligent & Robotic Systems, 108(3),
design optimization inspired by dingoes hunting strategies. Mathematical Problems in 1–16. https://doi.org/10.1007/s10846-023-01907-1
Engineering, 2021, 1–19. https://doi.org/10.1155/2021/9107547 Xie, L., Han, T., Zhou, H., Zhang, Z.-R., Han, B., & Tang, A. (2021). Tuna Swarm
Pereira, J. L. J., Francisco, M. B., Diniz, C. A., Antônio Oliver, G., Cunha, S. S., & Optimization: A novel swarm-based metaheuristic algorithm for global optimization.
Gomes, G. F. (2021). Lichtenberg Algorithm: A novel hybrid physics-based meta- Computational Intelligence and Neuroscience, 2021, 9210050. https://doi.org/
heuristic for global optimization. Expert Systems with Applications, 170, Article 10.1155/2021/9210050
114522. https://doi.org/10.1016/j.eswa.2020.114522 Xu, B., Heidari, A. A., Cai, Z., & Chen, H. (2023). Dimensional decision covariance colony
Połap, D., & Woźniak, M. (2021). Red Fox Optimization Algorithm. Expert Systems with predation algorithm: Global optimization and high− dimensional feature selection.
Applications, 166, Article 114107. https://doi.org/10.1016/j.eswa.2020.114107 Artificial Intelligence Review, 1–57. https://doi.org/10.1007/s10462-023-10412-8
Rahman, C. M. (2023). Group learning algorithm: A new metaheuristic algorithm. Neural Xue, J. K., & Shen, B. (2020). A novel swarm intelligence optimization approach:
Computing and Applications, 1–16. https://doi.org/10.1007/s00521-023-08465-5 Sparrow Search Algorithm. Systems Science & Control Engineering, 8(1), 22–34.
Rashedi, E., Nezamabadi-pour, H., & Saryazdi, S. (2009). GSA: A gravitational search https://doi.org/10.1080/21642583.2019.1708830
algorithm. Information Sciences, 179(13), 2232–2248. https://doi.org/10.1016/j. Yang, X. S. (2010). Engineering Optimization: An Introduction with Metaheuristic
ins.2009.03.004 Applications. John Wiley & Sons. http://www.researchgate.net/publication/259
Ray, G. C., Hufford, G. L., Overland, J. E., Krupnik, I., McCormick-Ray, J., Frey, K., & 603793_Engineering_Optimization_An_Introduction_with_Metaheuristic_Applicatio
Labunski, E. (2016). Decadal Bering Sea seascape change: Consequences for Pacific ns.
walruses and indigenous hunters. Ecological Applications, 26(1), 24–41. https://doi. Yang, Y., Chen, H., Heidari, A. A., & Gandomi, A. H. (2021). Hunger games search:
org/10.1890/15-0430 Visions, conception, implementation, deep analysis, perspectives, and towards
Rezvani, K., Gaffari, A., & Dishabi, M. R. E. (2023). The Bedbug Meta-heuristic Algorithm performance shifts. Expert Systems with Applications, 177, Article 114864. https://doi.
to Solve Optimization Problems. Journal of Bionic Engineering, 1–21. https://doi.org/ org/10.1016/j.eswa.2021.114864
10.1007/s42235-023-00356-8 Yuan, Y., Ren, J., Wang, S., Wang, Z., Mu, X., & Zhao, W. (2022). Alpine Skiing
Salehi-Amiri, A., Zahedi, A., Akbapour, N., & Hajiaghaei-Keshteli, M. (2021). Designing a Optimization: A new bio-inspired optimization algorithm. Advances in Engineering
sustainable closed-loop supply chain network for walnut industry. Renewable and Software, 170, Article 103158. https://doi.org/10.1016/j.advengsoft.2022.103158
Sustainable Energy Reviews, 141, 110821-. https://doi.org/10.1016/j. Zahedi, A., Salehi-Amiri, A., Hajiaghaei-Keshteli, M., & Diabat, A. (2021). Designing a
rser.2021.110821 closed-loop supply chain network considering multi-task sales agencies and multi-
Sandgren, E. (1990). Nonlinear integer and discrete programming in mechanical design mode transportation. Soft Computing, 25(5). https://doi.org/10.1007/s00500-021-
optimization. Journal of Mechanical Design, 112(2), 223–229. https://doi.org/ 05607-6
10.1115/1.2912596 Zahedi, A., Salehi-Amiri, A., Smith, N. R., & Hajiaghaei-Keshteli, M. (2021). Utilizing IoT
Seyyedabbasi, A., & Kiani, F. (2022). Sand Cat Swarm Optimization: A nature-inspired to design a relief supply chain network for the SARS-COV-2 pandemic. Applied Soft
algorithm to solve global optimization problems. Engineering with Computers, 1–25. Computing, 104, Article 107210. https://doi.org/10.1016/j.asoc.2021.107210
https://doi.org/10.1007/s00366-022-01604-x Zhang, C., & Wang, H. P. (1993). Mixed-discrete nonlinear optimization with simulated
Shehadeh, H. A. (2023). Chernobyl disaster optimizer (CDO): A novel meta-heuristic annealing. Engineering Optimization, 21(4), 277–291. https://doi.org/10.1080/
method for global optimization. Neural Computing and Applications, 35(15), 03052159308940980
10733–10749. https://doi.org/10.1007/s00521-023-08261-1 Zhong, C., Li, G., & Meng, Z. (2022). Beluga Whale Optimization: A novel nature-inspired
Shin, J. G., Kwon, O. H., & Ryu, C. (2008). Heuristic and metaheuristic spatial planning metaheuristic algorithm. Knowledge-Based Systems, 251, Article 109215. https://doi.
of assembly blocks with process schedules in an assembly shop using differential org/10.1016/j.knosys.2022.109215
evolution. Production Planning & Control, 19(6), 605–615. https://doi.org/10.1080/ Zitouni, F., Harous, S., Belkeram, A., & Hammou, L. E. B. (2022). The Archerfish Hunting
09537280802474941 Optimizer: A novel metaheuristic algorithm for global optimization. Arabian Journal
Sörensen, K. (2015). Metaheuristics—The metaphor exposed. International Transactions in for Science and Engineering, 47(2), 2513–2553. https://doi.org/10.1007/s13369-021-
Operational Research, 22(1), 3–18. https://doi.org/10.1111/itor.12001 06208-z
Stephen, S., Christu, D., & Dalvi, A. (2018). Design optimization of weight of speed Zitouni, F., Harous, S., & Maamri, R. (2021). The Solar System Algorithm: A novel
reducer problem through matlab and simulation using ansys. International Journal of metaheuristic method for global optimization. IEEE Access, 9, 4542–4565. https://
Mechanical Engineering and Technology, 9, 339–349. doi.org/10.1109/ACCESS.2020.3047912
Talatahari, S., Azizi, M., Tolouei, M., Talatahari, B., & Sareh, P. (2021). Crystal Structure
Algorithm (CryStAl): A metaheuristic optimization method. IEEE Access, 9,
71244–71261. https://doi.org/10.1109/ACCESS.2021.3079161

33