Comparative Analysis of Bio-inspired Routing

Protocols for MANET: Dragonfly, Honeybee,

and Bacteria Foraging Algorithms

Arnab Das

Department of BCA, Institute of Hotel and Restaurant Management, Kolkata,

India-700150

Abstract. Mobile Ad Hoc Networks (MANETs) are self-configuring

networks of mobile devices connected without any fixed infrastructure.
Efficient routing in MANETs is challenging due to dynamic topologies,
limited bandwidth, and energy constraints. Bio-inspired algorithms offer
promising solutions by mimicking natural behaviors to optimize rout-
ing performance. This paper provides a comparative analysis of three
bio-inspired algorithms—Dragonfly Algorithm (DA), Honeybee Algo-
rithm (HA), and Bacteria Foraging Algorithm (BFA)—in the context of
MANET routing protocols. The study evaluates these algorithms based
on performance metrics such as packet delivery ratio, end-to-end delay,
routing overhead, and energy consumption.

Keywords: Mobile Adhoc networks · Honey Bee · Dragonfly · Routing

· Throughput.

1 Introduction
1.1 A Subsection Sample
Mobile Ad Hoc Networks (MANETs) are groups of mobile wireless nodes that
form decentralized networks without any centralized control. The primary role
of MANETs is to identify and establish paths between nodes, enabling effi-
cient communication. MANETs are characterized by complex topology, energy
constraints, varying node density, multi-hop communication, and limited band-
width.
The enrollment in MANETs has grown to address these challenges, leveraging de-
sign approaches to solve problems more efficiently. However, there remains a need
for resilience and detailed numerical analysis of existing methods to effectively
manage these issues . Traditional methods often struggle with the complexity of
MANETs, prompting the exploration of new, nature-inspired techniques. These
bio-inspired methods offer straightforward solutions to complex problems that
conventional computational approaches find difficult to address.
In a MANET, nodes must forward packets to each other, either through direct
communication within their radio range or by using intermediate nodes as re-
lay points. Unlike systems that rely on a fixed infrastructure, MANETs operate
2 Arnab Das

autonomously, which has sparked significant interest due to their flexibility and
robustness in various applications.
This paper focuses on three bio-inspired algorithms for MANET routing:
Dragonfly Algorithm (DA): Inspired by the swarming behaviors of dragonflies,
DA simulates their alignment, cohesion, and separation mechanisms to balance
exploration and exploitation in dynamic network environments.
Honeybee Algorithm (HA): Mimicking the foraging behavior of honeybees, HA
involves scout bees exploring new routes and forager bees exploiting known
routes, dynamically balancing the discovery and use of routing paths.
Bacteria Foraging Algorithm (BFA): Based on the foraging behavior of E. coli
bacteria, BFA employs chemotaxis, reproduction, and elimination-dispersal pro-
cesses to adaptively find and maintain efficient routing paths.
The objective of this paper is to provide a comparative analysis of the Dragonfly
Algorithm (DA), Honeybee Algorithm (HA), and Bacteria Foraging Algorithm
(BFA) as routing protocols in MANETs. By evaluating these algorithms based
on performance metrics such as packet delivery ratio, end-to-end delay, routing
overhead, and energy consumption, this study aims to determine their suitability
and effectiveness in various network scenarios.
The remainder of the paper is structured as follows: Section 2 outlines the
methodology, including the simulation environment and performance metrics.
Section 3 presents the results and discussion, highlighting the comparative per-
formance of the three algorithms. Finally, Section 4 concludes the paper and
suggests directions for future research.

2 Literature review

The field of Mobile Ad Hoc Networks (MANETs) has seen significant advance-
ments, with various routing protocols developed to address the inherent chal-
lenges of dynamic topology, limited bandwidth, and energy constraints. Among
these, bio-inspired swarm algorithms have garnered considerable attention due
to their ability to provide robust and adaptive routing solutions. Recent re-
search has focused on leveraging the collective behavior of natural swarms to
optimize routing in MANETs. This section reviews the latest research on swarm
algorithm-based routing protocols, highlighting key examples.
1. [1] Ant Colony Optimization (ACO): Ant Colony Optimization (ACO) is in-
spired by the foraging behavior of ants, where ants deposit pheromones to mark
favorable paths, which other ants subsequently follow. This paper proposes an
enhanced ACO-based routing protocol that incorporates adaptive pheromone
evaporation rates and dynamic path updating mechanisms to improve route sta-
bility and reduce latency in MANETs.
2. [2] Particle Swarm Optimization (PSO): Particle Swarm Optimization (PSO)
mimics the social behavior of bird flocks or fish schools, where particles (solu-
tions) adjust their positions based on personal and collective experiences. The
study introduces a hybrid routing protocol that combines PSO with traditional
 Title Suppressed Due to Excessive Length 3

AODV (Ad hoc On-Demand Distance Vector) routing, enhancing route discov-
ery and maintenance processes, leading to improved packet delivery ratios and
reduced overhead.
3. [3] Honey Bee Colony Optimization (BCO): Bee Colony Optimization (BCO)
is inspired by the foraging behavior of honeybees. Scout bees explore new ar-
eas, and forager bees exploit known sources. This paper presents an adaptive
BCO-based routing protocol that dynamically adjusts the number of scout and
forager bees based on network conditions, enhancing route discovery and reduc-
ing end-to-end delay.
4. [4]Firefly Algorithm (FA): Firefly Algorithm (FA) is inspired by the flashing
behavior of fireflies, where fireflies are attracted to brighter ones, representing
better solutions. The proposed protocol uses the FA to optimize route selection
based on node attractiveness and link quality, resulting in higher packet delivery
ratios and lower energy consumption compared to conventional methods.
5.[5]Bat Algorithm (BA): Bat Algorithm (BA) is inspired by the echolocation
behavior of bats, which navigate and hunt using sound waves. This research in-
troduces a BA-based routing protocol that leverages echolocation principles to
dynamically adjust routing paths, enhancing network scalability and reliability.
6. [6] Dragonfly Algorithm (DA): The Dragonfly Algorithm (DA) simulates the
static and dynamic swarming behaviors of dragonflies. This study proposes a DA-
based routing protocol that optimizes route discovery and maintenance through
swarming behaviors, resulting in improved route stability and reduced routing
overhead.
7. [7]Cuckoo Search Algorithm (CSA): Cuckoo Search Algorithm (CSA) is in-
spired by the brood parasitism of some cuckoo species. It has been applied to
MANET routing to optimize path selection. This research introduces a CSA-
based routing protocol that optimizes route selection by balancing exploration
and exploitation, leading to higher packet delivery rates and reduced latency .

The literature reveals a diverse range of bio-inspired swarm algorithms that

have been adapted for MANET routing protocols, each leveraging unique natu-
ral behaviors to address the challenges of dynamic and decentralized networks.
These algorithms demonstrate significant potential in enhancing the efficiency,
reliability, and scalability of MANET routing, making them promising candi-
dates for future research and application.
The following sections of this paper will delve into the methodology for compar-
ing Dragonfly Algorithm (DA), Honeybee Algorithm (HA), and Bacteria Forag-
ing Algorithm (BFA)[9], followed by a detailed analysis of their performance in
various network scenarios.
4 Arnab Das

3 Motivation
3.1 Problem formulation
To determine the most efficient route in terms of packet delivery ratio, one-way
delay, and throughput, we formulated the routing problem and compared various
protocols. We then employed bio-inspired techniques, such as the Honeybee and
Dragonfly algorithms, to identify the best approach.

3.2 Objectives of the study

– To propose a protocol improvement that enhances the establishment of
routes between the source and destination.
– Utilizing both the Honeybee and Dragonfly algorithms, we implemented the
proposed enhancements and evaluated their impact on the current protocol
in terms of packet delivery ratio, end-to-end latency, throughput, and packet
loss.

4 Contribution
4.1 Honeybee Algorithm:
The Honeybee algorithm (HBA) [3] is fundamentally a population-based, nature-
inspired approach designed to tackle optimization problems by mimicking the
foraging behavior of honeybee swarms. Drawing from the principles observed
in real honeybee colonies—such as food search, defense, and relocation—this
approach organizes artificial bees into three distinct roles: working bees, on-
looker bees, and scout bees, each assigned specific tasks corresponding to their
roles. The algorithm operates within a search space, initializing the exploration
process with the deployment of scout bees, whose fitness is evaluated. Bees sur-
passing a predefined fitness threshold are selected, and their visited locations
are earmarked for communal exploration. Iterative strategies based on Swarm
Intelligence (SI) guide the algorithm’s progression, with termination conditions
carefully set to dictate the algorithm’s conclusion. In the context of ad hoc and
wireless sensor networks, Honeybee-based algorithms find significant application.

4.2 Dragonfly Algorithm:

The Dragonfly Algorithm is fascinating in how it mirrors the collective behavior
of beings in groups, which efficiently observe and perform tasks. This capacity
has evolved over centuries, refining the most optimal and economical habits.
Leveraging these natural behaviors to address our problems is both intriguing
and cost-effective. In the realm of science, this approach aligns with swarm in-
telligence (SI), which involves simulating the collective and social intelligence
observed in nature. Researchers in this field aim to decipher the inherent rules
 Title Suppressed Due to Excessive Length 5

governing interactions among individuals that lead to social intelligence. Without

a centralized control unit directing individuals, simulating the social behaviors
of the entire population allows for the discovery of fundamental rules among di-
verse entities. Among the various SI techniques, based on the foraging behavior
of E. coli bacteria, BFA optimizes routing by simulating chemotaxis, reproduc-
tion, and elimination-dispersal processes.

5 Proposed Dragonfly-Bee routing algorithm [10],[11],

[12],[13] : A successful meta-heuristic algorithm necessitates a harmonious inte-

gration of leveraging existing knowledge accumulated within the search strategy
while also exploring new facets of the optimal solution, potentially leading to
further enhancements. The Dragonfly Algorithm represents a contemporary op-
timization approach that faithfully emulates the lifestyle of dragonfly insects.
This algorithm exhibits significant potential to gather robust search samples by
considering multiple variables, as depicted in equation (1), thereby influencing
the equilibrium between exploitation and exploration (2).

∆t + 1 = (vV j + wW j + dDj + bBj + rRj) + ω∆Xt (1)

xt + 1 = Xt + ∆Xt + 1 (2)
The first equation represents a model that simulates the behavior of a swarm of
bees foraging for food. Here’s a breakdown of the symbols:
Segregation Volume (v): This represents the space each bee occupies while
searching.
Individual Segregation (Vj): This reflects how much each bee (jth) keeps its dis-
tance from others.
Coordination Weight (w): This signifies the importance of following other bees’
discoveries.
Individual Coordination Weight (Wj): This indicates how much each bee (jth)
relies on information from others.
Cohesion Weight (d): This represents the force that keeps the swarm together.
Individual Cohesion (Dj): This reflects how much each bee (jth) is attracted to
staying with the swarm.
Food Factor (b): This signifies the attractiveness of a food source. Individual
Food Source (Bj): This indicates the food source each bee (jth) is currently ex-
ploring.
Enemy Factor (r): This represents the danger or difficulty associated with a food
source.
Individual Enemy Role (Rj): This reflects each bee’s (jth) perception of the risk
involved.
Inertia Weight (w): This controls the balance between exploration and exploita-
tion (searching new areas vs. focusing on promising ones).
Iteration Number (t): This refers to the current stage of the search process.
6 Arnab Das

Original Location (X): This represents the starting point for each bee. Strengths
and Weaknesses of Existing Models:
The traditional Artificial Bee Colony (ABC) algorithm excels at finding good
solutions (local optima) by relying on two phases: employed bees and onlooker
bees. Employed bees share information about the food sources they’ve found,
while onlooker bees choose the most promising sources based on this informa-
tion.
However, the ABC algorithm primarily focuses on local search, which can some-
times trap the search in suboptimal areas, especially during the initial stages.
Additionally, the scouting phase, meant for global exploration, might not be as
effective in finding diverse food sources initially.
Introducing the Hybrid Dragonfly-Bee Algorithm:
To address the limitations of existing models, a new hybrid algorithm called
Dragonfly-Bee has been proposed. This algorithm combines aspects of the ABC
algorithm with the Dragonfly Algorithm, aiming for a better balance between
local and global search.
We’ll delve deeper into the specifics of the Dragonfly-Bee algorithm in the fol-
lowing section, including a detailed explanation of the onlooker bee phase (DA
stage).

Algorithm 1 Dragonfly-Bee algorithm

1: Input: Objective function f(x) and limitations
2: initiation
3: (i)initiation of parameters: M G(M aximumGeneration);FS (Food Sites Number);
Restrict; division of weight; weight coordination; alignment weight; food interaction
weight; opponent diversion weight;
4: (ii) Initiation of the population: population of dragonflies xj J = 1, 2. . .
F S), stepvectors∆xj(J = 1, 2...FS);
5: 3(iii) Set Pro = 0.1 and generation it = 0
6: test cases
7: While it ≤MG
8: for J = 1, 2. . . FS
9: if rand ≤Pro then
10: Dragonfly algorithm()
11: else
12: On-looker phase()
13: end for
14: it =it + 1
end while
15: The final phase
16: Output: the better solution than honeybee and dragonfly individually

Boosting Exploration in the Dragonfly-Bee Algorithm:

The Dragonfly-Bee algorithm introduces two key improvements to the tradi-
tional ABC algorithm:
 Title Suppressed Due to Excessive Length 7

Enhanced Employed Bee Search: The way employed bees search for food sources
is modified to explore a wider range of areas. This helps overcome the ABC al-
gorithm’s weakness in global search, ensuring it doesn’t get stuck in suboptimal
areas.
Dragonfly Operator Integration: During the initial update phase (employed bee
phase), the ABC algorithm’s standard operator is replaced with the ”Dragonfly
Operator” inspired by the Dragonfly Algorithm. This operator encourages ex-
ploration and helps the bees find diverse food sources.
The improved operator, called the Dragonfly-Bee step, is shown in Algorithm 2.
Algorithm 3 details the ”immobilized observer bee step” which comes later in
the process.
Identifying and Replacing Exhausted Food Sources:
After the Dragonfly-Bee process and the onlooker bee phase, the Dragonfly-Bee
algorithm checks if any food sources are depleted and should be abandoned.
Special counters are used to track how long it’s been since a source yielded a
good solution.
If a Dragonfly-Bee doesn’t find promising new solutions at a source for a certain
number of steps, the counter increases. When the counter reaches a predefined
limit, it’s time to replace the source.
A new food source is created using Equation (assuming xj is the discarded source)
and is then explored by a modified bee team (revised bee squad). This ensures
the search continues in promising new areas.

Algorithm 2 Dragonfly-Bee phase

1: Calculate all dragonflies’ objective values
2: Monitor the food supplies and the opponent
3: Monitor v, w, d, b, b, and r
4: Evaluate V, W, D, B, and R
5: Review surrounding range
6: if A dragonfly must have at least an adjacent dragonfly then
7: Analysis velocity vector of equation (1) and position vector of equation (2)
8: else
9: Analysis position vector of equation (2)
end if
10: Measure the objective function of the candidate solution and using a greedy method
of selection to choose the best possible one ;

Setting Up the Dragonfly-Bee Algorithm:

Before running the Dragonfly-Bee algorithm, we need to configure its parame-
ters. These parameters can be broadly categorized into three groups:
Control Parameters: These govern the overall behavior of the algorithm, like the
number of iterations or population size.
ABC Parameters: These are settings inherited from the traditional ABC algo-
rithm, such as the number of employed bees or onlooker bees.
 8 Arnab Das

Algorithm 3 On looker bee phase

1: Input: Place of a dragonfly xj
2: Choose from all xj’s higher accuracy values
3: For the chosen xj, measure the probability values p
4: for every dragonfly and honey bee, do
5: if probability value ≥ random(0, 1)thenf ollowagreedyselectiontoselectthebestanswer
6: end if
7: End for

DA Parameters: These are specific to the Dragonfly Operator and influence its
exploration behavior.
The Dragonfly-Bee algorithm builds upon existing settings from the honeybee
and dragonfly algorithms while introducing a new parameter: the probability
threshold (Pro). This value controls the balance between exploration and ex-
ploitation in the algorithm by influencing the Dragonfly-Bee mechanism and the
onlooker bee phase. In this particular implementation, the probability threshold
(Pro) is set to 0.1. This means there’s a 10% chance that the Dragonfly Opera-
tor will be used during the employed bee phase, promoting exploration alongside
exploitation.

6 Simulation
The above assessment in this analysis contains many critical success factors,
which are implemented in the MATLAB environment and discussed below:
Average End-to-End Delay:The average time taken for a packet to travel
from the source to the destination. This delay involves several lower latency de-
lays, along with all possible cacheing delays, waiting at the list of the router
interface, MAC re-transmission disruptions, propagation, and transfer time dur-
ing the latency of route discovery. The average
Pn E2E delay of data packets can be
calculated using the following formula: i−1 (RT i − ST i )/n, where RT i receives
packet time I, ST i is the total packet sent, and n is sent packet time i.
Average Throughput:The estimated throughput metric is the combination of
the successful data streams acquired over the total simulation time period. The
average throughput is calculated for each second in kilobits (kbps) and tests the
efficiency and effectiveness of the routing protocol when the destinations receive
data packets. The following
P Pformula is used to calculate the average throughput:
kt received at destinations 8
Average Throughput = stop time−start time × 1000
Packet Delivery Ratio: The metric of the packet delivery ratio indicates
the overall number of knowledge packets received by destinations, divided by
the entire number of knowledge packets sent by the source. This metric shows
how a protocol delivers packets from the source to the destination success-
fully. Good results, which reflect the completeness and correctness of the rout-
ing protocol, are indicated by a high packet delivery ratio. The packet deliv-
ery ratio is computed using the subsequent formula: Packet delivery Ratio =
 Title Suppressed Due to Excessive Length 9
P
P kt received at destinations
P
P kt sent by sources
× 100
The experimental results are presented in the given tables and graphs.

Table 1. The simulation environment parameters

Parameter Description Type

Number of Nodes 50 Node
Queue size 100 Packet
Simulation area 1670 × 970 m2
Routing protocols BFA and Dragonfly-Bee -
Mobility model Random way -
Packet size 256 Bytes
Transmission range 250 M
Type of traffic CBR -
Initial energy 100 Jules
Idle power consumption 0.05 mw
Initial energy 100 Jules
Transmission power consumption 1.35 mw
Receive power consumption 1.7 mw
Sleep power 0.001 mw

Table 2. End to end delay comparison between Dragonfly-Bee and Bacteria Foraging

Number of Node BFA Dragonfly-Bee

20 24.807 35.201
40 46.692 36.066
60 53.577 37.931
80 66.462 39.796
100 74.347 41.661
120 87.232 43.526
140 91.117 45.391
160 101.002 47.256
180 112.887 48.121
200 122.772 49.986

7 COMPARATIVE ANALYSIS OF PROPOSED AND

EXISTING WORK

This section mentions the contrast of current work with a novel mechanism[8].
The analogy was made to clarify the implied work quality. The PDR is a crucial
metric that indicates the reliability of the routing protocol. The results show
10 Arnab Das

Table 3. Throughput comparison between Dragonfly-Bee and Bacteria Foraging

Number of Node BFA Dragonfly-Bee

20 46.16 40.92944
40 43.70296 38.56638
60 41.24592 36.20332
80 38.78888 33.84026
100 36.33184 31.4772
120 33.8748 29.11414
140 31.41776 26.75108
160 28.96072 24.38802
180 26.50368 22.02496
200 24.04664 19.6619

Table 4. Packet delivery comparison between Dragonfly-Bee and Bacteria Foraging

Number of Node BFA Dragonfly-Bee

20 99.719 99.64827.807
40 99.656 99.604
60 99.593 99.56
80 99.53 99.516
100 99.487 99.472
120 99.404 99.428
140 99.396 99.384
160 99.278 99.34
180 99.299 99.296
200 99.272 99.252
 Title Suppressed Due to Excessive Length 11

Fig. 1. Comparison of proposed schemes with respect to End to End Delay

150 BFA Dragonfly-Bee

End to End Delay

100

50

20 60 100140180200
Number of Nodes

Fig. 2. Comparison of proposed schemes with respect to Throughput

50
Dragonfly-Bee BFA
Throughput

40

30

20
20 60 100140180200
Number of Nodes

Fig. 3. Comparison of proposed schemes with respect to Packet delivery(%)

BFA Dragonfly-Bee
Packet Delivery %

99.6

99.4

99.2
20 60 100140180200
Number of Nodes

that Dragonfly-BEE consistently achieves a high PDR due to its efficient explo-
ration and exploitation balance, where as BFA exhibits lower PDR compared to
DA and HA, primarily due to the overhead associated with chemotactic move-
ments and frequent reproduction cycles. End-to-End Delay. The E2E delay is
critical for time-sensitive applications. Dragonfly-BEE maintains low E2E de-
12 Arnab Das

lay, benefiting from its cohesive and aligned movements that quickly adapt to
network changes. On the other hand, BFA has higher E2E delay, attributed to
the iterative chemotaxis and reproduction processes that can introduce delays in
establishing routes. Routing throughput impacts network efficiency. Dragonfly-
Bee generates moderate routing overhead, balancing between exploration and
exploitation where as BFA has the highest routing overhead, primarily due to
frequent chemotactic movements and the need for maintaining bacterial popu-
lations.

8 Conclusion
The comparative analysis of bio-inspired routing protocols for MANETs reveals
that each algorithm has its strengths and weaknesses. Dragonfly-BEE Algorithm
excels in PDR and E2E delay, making it suitable for applications requiring high
reliability and low latency. Bacteria Foraging Algorithm (BFA) provides robust
adaptability to dynamic environments but at the cost of higher routing overhead
and energy consumption. Future research could explore hybrid approaches that
combine the strengths of these algorithms to enhance overall MANET perfor-
mance. The results indicate that the Dragonfly-BEE Algorithm is particularly
well-suited for applications requiring high reliability and low latency, making
it an excellent choice for MANET routing. The Bacteria Foraging Algorithm
(BFA) provides robust adaptability to dynamic environments but at the cost of
higher routing overhead and energy consumption.

8.1 Author Contribution

The sole author of this paper was responsible for the conceptualization, method-
ology, software, formal analysis, data curation, writing—original draft prepara-
tion, validation, resources, and visualization. The author has read and approved
the final version of the manuscript.

9 Funding
This research received no external funding.

10 Acknowledgements
The author would like to thank Institute of Hotel and restaurant management,
Kolakta for providing the resources and support necessary to conduct this re-
search.
 Title Suppressed Due to Excessive Length 13

11 Conflicts of Interest

The author declares no conflict of interest.

References
1. Mirjalili, S., Dragonfly algorithm: a new meta-heuristic optimization technique for
solving single-objective, discrete, and multi-objective problems. Neural Computing
and Applications, 27(4), 1053-1073, (2015).
2. Sharma, S.,& Purohit, G. N. Comparative analysis of bio-inspired algorithms for
optimizing MANET routing protocols. International Journal of Applied Engineering
Research, 10(3), 2471-2481,(2015).
3. Tamilarasan, S. A quantitative study and comparison of AODV, OLSR and TORA
routing protocols in MANET. International Journal of Computer Science Issues
(IJCSI), 9(1), 364.(2012)
4. Saranya, C. T., & Blossom, M. S. Comparison of AODV, OLSR and DSDV Routing
Protocol for MMANETs. Journal of Recent Research in Engineering and Technol-
ogy, pp1–6 Article ID J11601 ISSN (Online), 2349-2252.(2016).
5. Kaur, M., & Gangal, A. Comparative Analysis of Various Routing Protocol
in MANET. International Journal of Computer Applications, 118(8), 20766-
3207.(2015).
6. Aggarwal, N., Chohan, T. S., Singh, K., Vohra, R., & Bahel, S. Relative Analysis of
AODV & DSDV Routing Protocols for MANET based on NS2. In 2016 International
Conference on Electrical, Electronics, and Optimization Techniques (ICEEOT) (pp.
3500-3503). IEEE.(2016, March).
7. Raja, M. L., & Babo, C. D. S. S. Comparative study of reactive routing protocol
(AODV, DSR, ABR and TORA) in MANET. International Journal Of Engineering
And Computer Science, 2(03).(2013).
8. Chezhiyan, D. U. Measurement based analysis of reactive protocols in manet. In-
ternational Journal of Wired and Wireless Communications, 1(2).(2013).
9. Mubin, N., Ali, S. A., & Iqbal, K. Enhanced Bacterial Foraging Optimization
for Energy-Efficient MANETs. International Journal of Computer Applications,
177(28), 1-7,(2022).
10. Singh, C. A Review: Comparative Study of Routing Protocols for Adhoc Networks.
International Journal of Advanced Research in Computer and Communication En-
gineering, 4, 319-321.(2015).
11. D. Karaboga, B. Akay, and C. Ozturk, “Articial bee colony (ABC) optimization
algorithm for training feed-forward neural networks,” in Proc. Int. Conf. Modeling
Decisions Artif. Intell., Berlin, Germany: Springer,pp. 318329,2007.
12. W. A. H. M. Ghanem and A. Jantan, “A cognitively inspired hybridization of
articial bee colony and dragony algorithms for training multi-layer perceptrons,”
Cognit. Comput., vol. 10, no. 6, pp. 10961134, Dec. 2018.
13. S. Mirjalili, “Dragony algorithm: A new meta-heuristic optimization technique for
solving single-objective, discrete, and multi-objective problems,” Neural Comput.
Appl., vol. 27, no. 4, pp. 10531073,May 2016.