Cuckoo algorithm with great deluge local-search for feature selection problems

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 12, No. 4, August 2022, pp. 4315~4326
ISSN: 2088-8708, DOI: 10.11591/ijece.v12i4.pp4315-4326  4315
Journal homepage: http://ijece.iaescore.com
Cuckoo algorithm with great deluge local-search for feature
selection problems
Mutasem Khalil Alsmadi1
, Malek Alzaqebah2,3
, Sana Jawarneh4
, Sami Brini2,3
, Ibrahim
Al-Marashdeh1
, Khaoula Briki2,3
, Nashat Alrefai2,3
, Fahad Ali Alghamdi1
, Maen T. Al-Rashdan5
1
Department of Management Information Systems, College of Applied Studies and Community Service, Imam Abdulrahman Bin Faisal
University, Dammam, Saudi Arabia
2
Department of Mathematics, College of Science, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia
3
Basic and Applied Scientific Research Center, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia
4
Computer Science Department, The Applied College, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia
5
Faculty of Ccience and Information Technology, Jadara University, Irbid, Jordan
Article Info ABSTRACT
Article history:
Received Oct 16, 2020
Revised Mar 25, 2022
Accepted Apr 18, 2022
Feature selection problem is concerned with searching in a dataset for a set
of features aiming to reduce the training time and enhance the accuracy of a
classification method. Therefore, feature selection algorithms are proposed
to choose important features from large and complex datasets. The cuckoo
search (CS) algorithm is a type of natural-inspired optimization algorithms
and is widely implemented to find the optimum solution for a specified
problem. In this work, the cuckoo search algorithm is hybridized with a local
search algorithm to find a satisfactory solution for the problem of feature
selection. The great deluge (GD) algorithm is an iterative search procedure,
that can accept some worse moves to find better solutions for the problem,
also to increase the exploitation ability of CS. The comparison is also
provided to examine the performance of the proposed method and the
original CS algorithm. As result, using the UCI datasets the proposed
algorithm outperforms the original algorithm and produces comparable
results compared with some of the results from the literature.
Keywords:
Classification
Cuckoo search
Feature selection
Great deluge
Metaheuristic optimization
This is an open access article under the CC BY-SA license.
Corresponding Author:
Mutasem Khalil Alsmadi
Department of MIS, College of Applied Studies and Community Service, Imam Abdulrahman Bin Faisal
University, P.O. Box 1982, 31441, City of Dammam, Saudi Arabia
Email: mkalsmadi@iau.edu.sa
1. INTRODUCTION
Classification is considered as one of the machine learning tasks, which have been widely used
recently to categorize the data into classes [1]–[5]. Classification techniques predict the classes of the data
instances based on a given set of data fields (features). Using the original number of features may be time-
consuming and may mislead the classification process, so the methods for feature selection chose a minimum
set of features that lead to better learning accuracy and less computational cost. The methods for feature
selection are separated into three categories [6]: i) filter-based methods that use statistical approaches to
assess the correlation between features and the class, ii) wrapper-based methods: assess the selected features’
subset using a machine learning algorithm, and iii) embedded-based methods: combines the advantage of
wrapper methods and filter-based methods [6]–[8].
Feature selection is an nondeterministic polynomial (NP)-problem because of its high-dimensional
space [9]–[12] where the exhaustive search is unfeasible. To perform the feature selection task, an efficient
search algorithm is required. Swarm intelligence is a group of population-based algorithms that contain

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 4, August 2022: 4315-4326
4316
algorithms inspired by the social insects/animals’ behaviors, which are called nature-inspired optimization
algorithms [6], [13]–[17].
Several nature-inspired optimization algorithms were implemented for solving feature selection
problems using the wrapper-based method, since its simple, natural representation and efficient in global
search [6], [7], [18]–[22]. Particle swarm optimization mimics the behavior of birds and is utilized to find the
best set of features [8], [23]–[25]. Ant colony optimization simulates the ants’ behavior in searching for food
and it has been used for feature selection problems [26]–[32]. Cat swarm optimization is inspired by cats
searching for their prey [33], [34]. Grey wolf optimization algorithm depends on how the wolf pack behaves
(the hierarchy and hunting) [35], [36]. The behavior of genuine moths in looking for light sources is
mimicked by moth flame optimization algorithms [37]–[40].
In Yang and Deb [41] the authors presented the cuckoo search (CS) algorithm for continuous
optimization problems [42], [43], CS is based on the attractive cuckoo bird’s breeding method. CS algorithm
was effectively proposed to problems from different domains such as mobile robot navigation [44] and
reliability–redundancy allocation [45]. CS algorithm has some advantages compared with other nature-
inspired optimization algorithms such as it explores some elitism types. Also, in CS, randomness is more
useful as a move size, where it is heavy-tailed with any likely large move size. And, because there are fewer
parameters to tune than with genetic algorithms and particle swarm optimization, it may be easier to adapt to
a broader range of optimization problems [42], [46]. Similarly, CS has global optima achievement and rapid
convergence. A binary CS (BCS) proposed for feature selection problems in [46], [47]. But there is a
limitation in the CS algorithm, which is it has a slow convergence speed [47]. Modified CS algorithm with
rough sets is proposed by [48] for feature selection problem, In the modified version some cuckoo species
use the obligate brood parasitic behavior and some birds the Lévy flight behavior. The dimensionality of the
datasets required an efficient search algorithm to discover the optimum features’ subset for better prediction,
thus we propose a modified CS algorithm with great deluge algorithm (GD) as local search, to overcome the
slow convergence speed of the CS algorithm along with avoiding the CS algorithm getting trapped in local
optima.
This paper is organized: the cuckoo search algorithm is presented in section 2 along with the
proposed approaches and great deluge algorithm. The results with some discussions are listed in section 3.
Comparison of CS algorithm with other results from the literature are stated in section 4. The conclusion and
future research are stated in section 5.
2. METHOD
The methodology presented here contains information about the original CS algorithm. Followed by
the modification of CS for feature selection problem and description of great deluge as a local-search
algorithm. Finally, the details of CS with great deluge algorithm.
2.1. Cuckoo search algorithm (CS)
CS algorithm was firstly proposed in [41]. The behavior of force brood parasitic for cuckoo
inspired the authors in [41] to develop a CS algorithm. This behavior starts from the cuckoo laying the eggs
in another small bird’s nest (host), normally the cuckoo’s eggs hatch before the eggs of the host, then the
cuckoo chick discovers an outlandish egg and decided to through other eggs. The representation of the CS
algorithm is: the population is represented by the nest, the solutions in the population are represented by
eggs and the new solutions that are produced using Levy-flight are the cuckoo’s eggs. Then the new solution
is compared to the other solutions and the best solutions are replaced by the worst solutions. The three main
rules in the CS algorithm are stated [41]: i) each cuckoo puts one egg in the random nest, ii) the higher
quality nests are set aside and considered for further improvement, and iii) the number of nests is
predetermined, for each nest, the cuckoo chick finds an outlandish egg using the probability between 0
and 1. Then the host chooses to abandon the nest or throw the egg.
The levy-flight presented in (1) is used by the CS algorithm for creating the new solutions.
𝑥𝑖+1 = 𝑥𝑖+α⊕Lvy(λ) (1)
The produced new solution for (cuckoo i) is xi, α is greater than 0, which represents the move size, λ
represents a constant of distribution of levy. Usually, the left side term in (1) denotes the random move
where the next location is based on the current location and the left term in (1) is the probability of a
transition, ⊕ representing the entry-wise multiplication. The move size here is multiplied by a number
chosen at random with a distribution of levy.

Int J Elec & Comp Eng ISSN: 2088-8708 
Cuckoo algorithm with great deluge local-search for feature selection problems (Mutasem Khalil Alsmadi)
4317
The random move using levy-flight is efficient for exploring all search regions, using its move
length is longer. Levy-flights allow a random move, and the random move length is given by a levy
distribution as shown in (2) [41].
𝐿𝑣𝑦~𝑢 = 𝑡 − 𝜆, (1 < 𝜆 ≤ 3 (2)
The process of the CS algorithm is presented in Figure 1, firstly the algorithm starts with initializing the
population with the number of host nests, after that in every iteration a randomly selected cuckoo (solution)
for generating a new solution using levy-flights.
Set Nests:the number of nests
Initialize the Population with Nests;
Set CSNiter: the number of Iterations;
i=0;
for i=0 to CSNiter do:
Co= a randomly selected cuckoo
Co*= levy-flight(Co);//produce a new solution for Co
FunCo*=FitnessFunction(Co*);
RN=Random nest among Nests
FunN= FitnessFunction(Co);
if FunCo* < FunN
RN = Co*;// Replace the solution
endif
remove the worst nests and replace new nests using levy-flight
i.e. The solutions of fraction pa∈[0,1] are abandon
output the best nests;
endwhile
Figure 1. The CS algorithm Pseudo-code [41]
2.2. Great deluge algorithm
Among algorithms that are based on water behavior [49], the great deluge (GD) algorithm was
firstly proposed by Dueck in 1993 [50], GD uses an acceptance criterion for accepting the neighbor solutions.
GD simulates the hill climber path in a great deluge while trying to maintain his feet dry. GD accepts the
neighbor solutions with worse objective value based on the water level (a boundary value). The level value
starts reducing with the decay rate during the search process. Reducing the value of level encourages the
working solution to consistently reduce till convergence.
The whole process of the GD algorithm is represented in Figure 2, the algorithm starts with
initializing the parameters then the iterative process starts, in every iteration, producing k neighboring
solutions from the input (current) solution (Co), line-8 Figure 2. The produced neighboring solution is
accepted if it’s better than the current solution or less than or equal to the water level (boundary), this
condition helps the GD to avoid getting trapped in local optima.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Function GD(Co): The input solution
Initialize K:# of neighborhood solutions
Initialize rain-Speed
Initialize GDIter: # of iterations:
water-Level ← f(Co)
decay-Rate ← f(Co) × rainSpeed/GDIter
i=0
while(i< GDIter)
Co* ← produce-Neighbour(Co,K)
if f(Co*) < f(Co) or f(Co) ≤ water-Level
Co ← Co*//accept the solution
endif
water-Level ← water-Level – decay-Rate
i++
endwhile
output Co
Figure 2. Pseudocode of the great deluge algorithm [51]

 ISSN: 2088-8708
4318
2.3. Cuckoo search with great deluge algorithms
CS algorithm in this section is utilized to select the best subset of features. The solution can be
represented for the feature selection problem as a matrix with size N containing 0 and 1, where N is the
whole features number in the given dataset, 0 indicates that the feature is not chosen, while 1 indicates that it
is chosen. CS algorithm disuse solutions based on a fraction, and produces a new solution, at an early stage of
the CS algorithm process, disusing the solution may be time-consuming and solutions didn’t improve, and
not enough iterations left to start improving a new solution, so an updating strategy before desertion the
solutions are required to improve the solutions by accepting the worse neighbor solutions.
The levy-flight is used by the CS algorithm to produce a new solution, we propose updating a
strategy that uses the great deluge algorithm with two neighborhood strategies, to avoid the CS algorithm
from getting stuck in the local optima and to speed up the convergence. Figure 3 represents the process of
cuckoo search algorithm with great deluge algorithm. The neighborhood strategies can be explained [52].
Let’s consider the solution is Co=[0, 0, 1, 1, 0, 1, 1, 0, 1, 0], so the neighborhood strategies are: i) move
neighborhood: chooses a feature at random and move its position to a new random position and ii) swap
neighborhood: chooses two features at random and swap values.
Figure 3. The process of CS algorithm with GD
In feature selection, two objectives should be taken into account to produce a good solution for the
problem, where the accuracy should be maximized as much as possible with minimizing the number of
selected features. thus, the k-nearest neighbor classifier (KNN) [53], used to produce the mean accuracy
using 10-fold-cross-validation [54], and the input features are given by the algorithm as a solution. So, the
Start
Initialize the population
with Nst host nests
Set i=0, Niter
Set pa [0,1]
Evaluate the Population
Co= select_random_cuckoo
Co*= GD(Co) See figure 2
FunC*=Evaluatefitness(Co*);
RN=random_nest()
FunRN=FitnessFunction(Co);
FunCo* > FunN RN = Co*
The solutions of fraction pa [0,1]
are abandon
save the best nests
Find the current best nests
i=i+1
End
iter<max_iter
yes
No
Final Solution &
Accuracy
build new nests using levy_flights
yes
No

4319
objective function (OF) in equation 3 is considered both objectives (maximize accuracy while reducing the
number of features selected) [53].
OF = 𝛼𝐸 + 𝛽 |
𝑆
𝑁
| (3)
Where the value of α is a parameter between 0 and 1 and β=1–α, E is the rate of error given by the KNN
classifier. S is the selected features’ number and N is the features' total number.
3. RESULTS AND DISCUSSION
The performance of this work is tested in this section using 9 UCI datasets which are used in several
well-confirmed research. These datasets are presented in Table 1 [55]. Experimental results based on
different values of parameters show the CS algorithm's final parameter settings as presented in Table 2. In the
original CS algorithm, parameters pa, , and  initialized firstly based on [56]. The findings of this study
have been implemented using a personal computer with the specifications: Intel i5-2.30 GHz Processor and
RAM of 8.0 GB. And, the results are conducted over 10 runs. The datasets are split into 80 training and 20
testings [55].
Table 1. UCI datasets used
Dataset Features Instances
1 German 20 1000
2 Breastcancer 9 699
3 Spect 22 267
4 Krvskp 36 3196
5 Ionosphere 34 351
6 Sonar 60 208
7 Lymphography 148 18
8 Tic-tac-toe 9 958
9 Wdbc 30 569
Table 2. Final parameters settings for CS algorithm
Algorithm Parameter Name Value
CS Pa 0.3
a 1
 1.5
Size of population 10
CS Iterations (CSNiter) 100
GD GD Iterations (GDIter) 100
#neighborhood solutions (K) 2
rain-Speed 0.5
3.1. Comparison between CS algorithm and CS with local search (CS_GD)
The results of the original CS algorithm presented in this section are evaluated and compared with
the proposed approach (CS_GD) to show its effectiveness. Table 3 shows the results, where the best results
for each dataset are represented by bold font. These two algorithms are compared based on the testing mean
accuracy in 10 folds cross-validation, the average accuracy of 10 runs, the selected features, and the time in
seconds taken to finish the process.
Table 3. Results comparison between CS algorithm and the proposed CS_GD
Dataset CS CS_GD
Accuracy Average
Accuracy
#Features
Selected
Average
Time
Accuracy Average
Accuracy
#Features
Selected
Average
Time
German 74.0 70.8 7 35.7 75.5 72.5 9 48.8
Breastcancer 92.9 92.9 3 15.3 92.9 92.9 3 28.0
Spect 83.3 75.7 16 10.2 83.3 76.9 11 15.1
Krvskp 96.6 94.7 14 161.6 97.5 95.1 10 232.5
Ionosphere 85.9 82.8 7 10.5 87.3 83.4 6 16.4
Sonar 88.1 76.9 20 9.2 90.5 78.3 11 14.2
Lymphography 80.0 69.7 5 6.8 80.0 72.7 4 10.5
Tic-tac-toe 89.1 89.1 9 16.5 89.1 89.1 9 30.7
Wdbc 92.1 91.0 8 13.1 93.0 90.1 6 23.9

 ISSN: 2088-8708
4320
The CS_GD algorithm presented better accuracies for 5 datasets and produces 4 similar accuracies
compared with the CS algorithm, and the results show that CS_GD 8 datasets have fewer selected features.
Based on the average computation time represented in Table 3 shows that the CS_GD algorithm needs
slightly more time to complete the process, nonetheless this extra time is worth it to produce better results.
The behavior of the CS_GD algorithm is presented in Figure 4 for Lymphography and German datasets,
where the objective function (OF) value is presented (3), in Figure 4, the number of iterations is represented
by the x-axis and the y-axis represents the objective function.
Figure 4. The convergence behavior of the current solution of CS and CS_GD
Using the great deluge algorithm, the solution is accepted in some cases based on the water level, so
the worst solution is accepted in some iterations to which helps the algorithm to escape from getting trapped
in local the optima and getting better solutions that improve the objective function, also GD speed up the
convergence behavior. Boxplots of accuracies produced by the CS and CS GD algorithms are compared and
exhibited in Figures 5 and 6 to investigate the reliability and stability of the findings. Each box shows the
median which is represented by the middle line in the box, while the top and bottom lines represent the
minimum and maximum values, respectively. The box plots for breast cancer and tic-tac-toe datasets show
the middle, top, and bottom lines as one line, which means that a similar result is represented for all runs, but
other box plots illustrate that the variance of maximum and minimum values in most of the datasets are as
acceptable and small, which represents the reliability and stability of the results.
Figure 5. Boxplots of CS algorithm for all datasets

4321
Figure 6. Boxplots of GD_CS algorithm for all datasets
3.2. Comparison between CS_GD and other nature-inspired algorithms
Three nature-inspired algorithms (bat algorithm (BAT), particle swarm optimization (PSO), and
firefly optimization algorithm (FFO)) are compared with the superior algorithm from the previous section
(CS_GD), The comparison is represented in Table 4, based on the average accuracy produced by 10
independent runs and the selected features’ number. As shown in Table 4 the average accuracies for CS_GD
outperform the PSO, BAT, and FFO algorithms for 7 datasets out of 9 datasets, 2 of them are similar; for the
Lymphography dataset, it has the same average accuracy in CS_GD and BAT algorithms and for Tic-tac-toe
have also the same average accuracy between CS_GD and FFO algorithms. Figure 7 represents the visual
results using the column chart to view the differences between the algorithms.
Table 4. Results of CS_GD, PSO, BAT and FFO algorithms
Dataset CS_GD PSO BAT FFO
Average
Accuracy
#Features
Selected
Average
Accuracy
#Features
Selected
Average
Accuracy
#Features
Selected
Average
Accuracy
#Features
Selected
German 72.5 9 68.5 16 70.1 8 71.5 9
Breastcancer 92.9 3 95.6 4 95.1 2 93.1 3
Spect 76.9 11 73.9 24 74.6 7 74.1 18
Krvskp 95.1 10 89.3 21 62.8 8 94.4 15
Ionosphere 83.4 6 80.4 18 82.5 10 81.3 10
Sonar 78.3 11 78.1 40 74.8 16 76.4 27
Lymphography 72.7 4 71.7 9 72.7 7 68.3 4
Tic-tac-toe 89.1 9 81.5 5 66.5 1 89.1 9
Wdbc 90.1 6 92.1 17 90.3 8 91.1 10
Figure 7. The comparison between CS_GD and other nature-inspired algorithms

 ISSN: 2088-8708
4322
The t-test is used to assess the significance of the acquired findings by calculating the difference
between the means of two groups. Table 5 provide the p-values that were obtained after applying the t-test
for CS_GD and other algorithms' average accuracy results. These statistical tests show that the observed
improvements and differences are significant. Where the detected differences between the CS_GD and FFO
are significant (p-value≤0.05) for 6 out of 8 datasets, the same number for the PSO algorithm.
Table 5. T-test of CS_GD with other algorithms
Dataset PSO BAT FFO
German 0.0070 0.0493 0.0120
Breastcancer 4.82E-05 0.0024 0.0839
Spect 0.2232 0.2799 0.0070
Krvskp 0.0072 4.18E-09 0.1527
Ionosphere 0.0188 0.2300 0.0204
Sonar 0.4504 0.1452 0.0239
Lymphography 0.3610 0.5000 0.0136
Tic-tac-toe 0.0009 3.81E-08 --
Wdbc 0.0072 0.4181 0.0805
3.3. Comparison of CS GD algorithm with other results from the literature
For the nine datasets studied in this work, the comparison between the best results achieved using
the CS_GD method and the best-known solutions from the literature is presented in this section. Table 6
compares the CS_GD algorithm with the most well-known findings of other algorithms from the literature.
When assessing the algorithms’ performance, accuracy is considered the primary goal. The highest level of
precision is shown in Table 6. As seen in Table 6, the CS_GD algorithm outperformed the other literature
results in 5 out of 9 datasets in terms of accuracy and has a comparable result with other comparators.
Table 6. Results of CS_GD Algorithm compared with some literature results
Dataset CS_GD Literature results Taken From
German 75.5 81.50 [24]
Breast cancer 92.9 98.00 [57]
Spect 83.3 82.60 [57]
Krvskp 97.5 96.80 [57]
Ionosphere 87.3 79.8 [58]
Sonar 90.5 86.70 [59]
Lymphography 80.0 85.30 [60]
Tic-tac-toe 89.1 80.80 [57]
WDBC 93.0 97.00 [24]
4. CONCLUSION AND FUTURE WORK
In this work, a hybridized CS algorithm with GD algorithm was introduced for feature selection
problem. Where, two objectives were considered for the feature selection problem, minimizing the number of
selected features and maximizing the prediction accuracy as possible. Thus; to achieve these objectives an
effective algorithm is necessary to be used to find an optimum solution of the problem. CS algorithm needs
to be modified to improve the convergence speed of the algorithm and to produce good results, thus GD
algorithm is proposed to enhance the solutions and to provide faster convergence to CS, due to the ability of
the GD algorithm to accept some worse moves to find better solutions. Using nine UCI datasets the
efficiency of the algorithm was exposed, the proposed method effectively finds good solutions compared
with some other nature-inspired algorithms and other comparable state-of-the-art methods. Our future work is
to find an automatic parameter tuning approach to set the parameters used and to investigate the proposed
algorithm in other domains.
ACKNOWLEDGMENTS
Many thanks to the Deanship of Scientific Research at the Imam Abdulrahman Bin Faisal
University. Where this research is funded by Imam Abdulrahman Bin Faisal University, under grant number
2019-416-ASCS.

4323
REFERENCES
[1] I. H. Sarker, “Machine learning: algorithms, real-world applications and research directions,” SN Computer Science, vol. 2, no. 3,
May 2021, doi: 10.1007/s42979-021-00592-x.
[2] M. K. Alsmadi, M. Tayfour, R. A. Alkhasawneh, U. Badawi, I. Almarashdeh, and F. Haddad, “Robust features extraction for
general fish classification,” International Journal of Electrical and Computer Engineering (IJECE), vol. 9, no. 6, pp. 5192–5204,
Dec. 2019, doi: 10.11591/ijece.v9i6.pp5192-5204.
[3] M. Wang, Z. Wei, M. Jia, L. Chen, and H. Ji, “Deep learning model for multi-classification of infectious diseases from
unstructured electronic medical records,” BMC Medical Informatics and Decision Making, vol. 22, no. 1, Dec. 2022, doi:
10.1186/s12911-022-01776-y.
[4] B. Charbuty and A. Abdulazeez, “Classification based on decision tree algorithm for machine learning,” Journal of Applied
Science and Technology Trends, vol. 2, no. 01, pp. 20–28, 2021.
[5] H. Wang, R. Czerminski, and A. C. Jamieson, “Neural networks and deep learning,” in The Machine Age of Customer Insight,
Emerald Publishing Limited, 2021, pp. 91–101.
[6] B. H. Nguyen, B. Xue, and M. Zhang, “A survey on swarm intelligence approaches to feature selection in data mining,” Swarm
and Evolutionary Computation, vol. 54, May 2020, doi: 10.1016/j.swevo.2020.100663.
[7] S. Fong, S. Deb, X.-S. Yang, and J. Li, “Feature selection in life science classification: metaheuristic swarm search,” IT
Professional, vol. 16, no. 4, pp. 24–29, Jul. 2014, doi: 10.1109/MITP.2014.50.
[8] L. M. Abualigah, A. T. Khader, and E. S. Hanandeh, “A new feature selection method to improve the document clustering using
particle swarm optimization algorithm,” Journal of Computational Science, vol. 25, pp. 456–466, Mar. 2018, doi:
10.1016/j.jocs.2017.07.018.
[9] A. A. Albrecht, “Stochastic local search for the feature set problem, with applications to microarray data,” Applied Mathematics
and Computation, vol. 183, no. 2, pp. 1148–1164, Dec. 2006, doi: 10.1016/j.amc.2006.05.128.
[10] G. Zhang, P. Yu, J. Wang, and C. Yan, “Feature selection algorithm for high-dimensional biomedical data using information gain
and improved chemical reaction optimization,” Current Bioinformatics, vol. 15, no. 8, pp. 912–926, Jan. 2021, doi:
10.2174/1574893615666200204154358.
[11] J. Li, S. Fong, R. K. Wong, R. Millham, and K. K. L. Wong, “Elitist binary wolf search algorithm for heuristic feature selection in
high-dimensional bioinformatics datasets,” Scientific Reports, vol. 7, no. 1, Dec. 2017, doi: 10.1038/s41598-017-04037-5.
[12] A. J. Tallon-Ballesteros, L. Correia, and S. Fong, “Hybrid scatter and ant search feature subset selection: applications in
classification problems,” in 2021 Second International Conference on Intelligent Data Science Technologies and Applications
(IDSTA), Nov. 2021, pp. 150–153, doi: 10.1109/IDSTA53674.2021.9660823.
[13] M. Alzaqebah, S. Jawarneh, H. M. Sarim, and S. Abdullah, “Bees algorithm for vehicle routing problems with time windows,”
International Journal of Machine Learning and Computing, vol. 8, no. 3, pp. 236–240, Jun. 2018, doi:
10.18178/ijmlc.2018.8.3.693.
[14] M. Alzaqebah, S. Abdullah, and S. Jawarneh, “Modified artificial bee colony for the vehicle routing problems with time
windows,” SpringerPlus, vol. 5, no. 1, Dec. 2016, doi: 10.1186/s40064-016-2940-8.
[15] W. Korani and M. Mouhoub, “Review on nature-inspired algorithms,” Operations Research Forum, vol. 2, no. 3, Sep. 2021, doi:
10.1007/s43069-021-00068-x.
[16] J. Zhao, D. Tang, Z. Liu, Y. Cai, and S. Dong, “Spherical search optimizer: a simple yet efficient meta-heuristic approach,”
Neural Computing and Applications, vol. 32, no. 13, pp. 9777–9808, Jul. 2020, doi: 10.1007/s00521-019-04510-4.
[17] M. K. Alsmadi et al., “An enhanced particle swarm optimization for ITC2021 sports timetabling,” Computers, Materials and
Continua, vol. 72, no. 1, pp. 1995–2014, 2022, doi: 10.32604/cmc.2022.025077.
[18] M. Alzaqebah, S. Jawarneh, R. M. A. Mohammad, M. K. Alsmadi, and I. ALmarashdeh, “Improved multi-verse optimizer feature
selection technique with application to phishing, spam, and denial of service attacks,” International Journal of Communication
Networks and Information Security (IJCNIS), vol. 13, no. 1, pp. 76–81, 2021.
[19] M. Sharawi, H. M. Zawbaa, E. Emary, H. M. Zawbaa, and E. Emary, “Feature selection approach based on whale optimization
algorithm,” in 2017 Ninth International Conference on Advanced Computational Intelligence (ICACI), Feb. 2017, pp. 163–168,
doi: 10.1109/ICACI.2017.7974502.
[20] R. A. Khurma, I. Aljarah, A. Sharieh, and S. Mirjalili, “EvoloPy-FS: an open-source nature-inspired optimization framework in
python for feature selection,” in Evolutionary machine learning techniques, 2020, pp. 131–173.
[21] D. Gupta, A. K. Ahlawat, A. Sharma, and J. J. P. C. Rodrigues, “Feature selection and evaluation for software usability model
using modified moth-flame optimization,” Computing, vol. 102, no. 6, pp. 1503–1520, Jun. 2020, doi: 10.1007/s00607-020-
00809-6.
[22] R. Guha, B. Chatterjee, S. K. Khalid Hassan, S. Ahmed, T. Bhattacharyya, and R. Sarkar, “Py_FS: A python package for feature
selection using meta-heuristic optimization algorithms,” in Computational Intelligence in Pattern Recognition, 2022, pp. 495–
504.
[23] Y. Lu, M. Liang, Z. Ye, and L. Cao, “Improved particle swarm optimization algorithm and its application in text feature
selection,” Applied Soft Computing, vol. 35, pp. 629–636, Oct. 2015, doi: 10.1016/j.asoc.2015.07.005.
[24] M. Alzaqebah et al., “Hybrid feature selection method based on particle swarm optimization and adaptive local search method,”
International Journal of Electrical and Computer Engineering (IJECE), vol. 11, no. 3, pp. 2414–2422, Jun. 2021, doi:
10.11591/ijece.v11i3.pp2414-2422.
[25] Y. Liu, G. Wang, H. Chen, H. Dong, X. Zhu, and S. Wang, “An improved particle swarm optimization for feature selection,”
Journal of Bionic Engineering, vol. 8, no. 2, pp. 191–200, Jun. 2011, doi: 10.1016/S1672-6529(11)60020-6.
[26] H. M. Rais and T. Mehmood, “Dynamic ant colony system with three level update feature selection for intrusion detection,”
International Journal of Network Security, vol. 20, no. 1, pp. 184–192, 2018.
[27] H. Arafat, R. M. Elawady, S. Barakat, N. M. Elrashidy, and W. S. Email, “Using rough set and ant colony optimization in feature
selection,” International Journal of Emerging Trends and Technology in Computer Science, vol. 2, no. 1, pp. 148–155, 2013.
[28] H. Liang, Z. Wang, and Y. Liu, “A new hybrid ant colony optimization based on brain storm optimization for feature selection,”
IEICE Transactions on Information and Systems, vol. E102.D, no. 7, pp. 1396–1399, Jul. 2019, doi:
10.1587/transinf.2019EDL8001.
[29] M. Paniri, M. B. Dowlatshahi, and H. Nezamabadi-pour, “Ant-TD: Ant colony optimization plus temporal difference
reinforcement learning for multi-label feature selection,” Swarm and Evolutionary Computation, vol. 64, Jul. 2021, doi:
10.1016/j.swevo.2021.100892.
[30] P. Shunmugapriya and S. Kanmani, “A hybrid algorithm using ant and bee colony optimization for feature selection and
classification (AC-ABC Hybrid),” Swarm and Evolutionary Computation, vol. 36, pp. 27–36, Oct. 2017, doi:

 ISSN: 2088-8708
4324
10.1016/j.swevo.2017.04.002.
[31] H. Peng, C. Ying, S. Tan, B. Hu, and Z. Sun, “An improved feature selection algorithm based on ant colony optimization,” IEEE
Access, vol. 6, pp. 69203–69209, 2018, doi: 10.1109/ACCESS.2018.2879583.
[32] B. Chen, L. Chen, and Y. Chen, “Efficient ant colony optimization for image feature selection,” Signal Processing, vol. 93, no. 6,
pp. 1566–1576, Jun. 2013, doi: 10.1016/j.sigpro.2012.10.022.
[33] K.-C. Lin, K.-Y. Zhang, Y.-H. Huang, J. C. Hung, and N. Yen, “Feature selection based on an improved cat swarm optimization
algorithm for big data classification,” The Journal of Supercomputing, vol. 72, no. 8, pp. 3210–3221, Aug. 2016, doi:
10.1007/s11227-016-1631-0.
[34] K.-C. Lin, Y.-H. Huang, J. C. Hung, and Y.-T. Lin, “Modified cat swarm optimization algorithm for feature selection of support
vector machines,” in Lecture Notes in Electrical Engineering, vol. 301, Springer Netherlands, 2014, pp. 329–336.
[35] C. Shen and K. Zhang, “Two-stage improved Grey Wolf optimization algorithm for feature selection on high-dimensional
classification,” Complex and Intelligent Systems, Jul. 2021, doi: 10.1007/s40747-021-00452-4.
[36] D. Jitkongchuen and P. Phaidang, “Grey wolf algorithm with borda count for feature selection in classification,” in 2018 3rd
International Conference on Control and Robotics Engineering (ICCRE), Apr. 2018, pp. 238–242, doi:
10.1109/ICCRE.2018.8376472.
[37] R. A. Khurma, I. Aljarah, and A. Sharieh, “Rank based moth flame optimisation for feature selection in the medical application,”
in 2020 IEEE Congress on Evolutionary Computation (CEC), Jul. 2020, pp. 1–8, doi: 10.1109/CEC48606.2020.9185498.
[38] A. Abuhamdah, M. Alzaqebah, S. Jawarneh, A. Althunibat, and M. Banikhalaf, “Moth optimisation algorithm with local search
for the permutation flow shop scheduling problem,” International Journal of Computer Applications in Technology, vol. 65, no. 3,
2021, doi: 10.1504/IJCAT.2021.116008.
[39] M. Alzaqebah, N. Alrefai, E. A. E. Ahmed, S. Jawarneh, and M. K. Alsmadi, “Neighborhood search methods with moth
optimization algorithm as a wrapper method for feature selection problems,” International Journal of Electrical and Computer
Engineering (IJECE), vol. 10, no. 4, pp. 3672–3684, Aug. 2020, doi: 10.11591/ijece.v10i4.pp3672-3684.
[40] M. Yarlagadda, K. Gangadhara Rao, and A. Srikrishna, “Frequent itemset-based feature selection and rider moth search algorithm
for document clustering,” Journal of King Saud University-Computer and Information Sciences, vol. 34, no. 4, pp. 1098–1109,
Apr. 2022, doi: 10.1016/j.jksuci.2019.09.002.
[41] X. S. Yang and S. Deb, “Cuckoo search via Lévy flights,” in 2009 World Congress on Nature and Biologically Inspired
Computing, NABIC 2009-Proceedings, 2009, pp. 210–214, doi: 10.1109/NABIC.2009.5393690.
[42] M. S. Al-Batah, S. Mrayyen, and M. Alzaqebah, “Arabic sentiment classification using MLP network hybrid with naive Bayes
algorithm,” Journal of Computer Science, vol. 14, no. 8, pp. 1104–1114, Aug. 2018, doi: 10.3844/jcssp.2018.1104.1114.
[43] X. S. Yang and S. Deb, “Engineering optimisation by cuckoo search,” International Journal of Mathematical Modelling and
Numerical Optimisation, vol. 1, no. 4, 2010, doi: 10.1504/IJMMNO.2010.035430.
[44] P. K. Mohanty and D. R. Parhi, “Cuckoo search algorithm for the mobile robot navigation,” in Lecture Notes in Computer Science
(including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 8297, no. 1, Springer
International Publishing, 2013, pp. 527–536.
[45] G. Kanagaraj, S. G. Ponnambalam, and N. Jawahar, “A hybrid cuckoo search and genetic algorithm for reliability-redundancy
allocation problems,” Computers and Industrial Engineering, vol. 66, no. 4, pp. 1115–1124, Dec. 2013, doi:
10.1016/j.cie.2013.08.003.
[46] X.-S. Yang, “Bat algorithm and cuckoo search: a tutorial,” in Studies in Computational Intelligence, vol. 427, Springer Berlin
Heidelberg, 2013, pp. 421–434.
[47] D. Rodrigues et al., “BCS: A binary Cuckoo search algorithm for feature selection,” in 2013 IEEE International Symposium on
Circuits and Systems (ISCAS2013), May 2013, pp. 465–468, doi: 10.1109/ISCAS.2013.6571881.
[48] M. A. El Aziz and A. E. Hassanien, “Modified cuckoo search algorithm with rough sets for feature selection,” Neural Computing
and Applications, vol. 29, no. 4, pp. 925–934, Feb. 2018, doi: 10.1007/s00521-016-2473-7.
[49] B. A. Aldeeb et al., “Hybrid intelligent water drops algorithm for examination timetabling problem,” Journal of King Saud
University-Computer and Information Sciences, Jun. 2021, doi: 10.1016/j.jksuci.2021.06.016.
[50] G. Dueck, “New optimization heuristics,” Journal of Computational Physics, vol. 104, no. 1, pp. 86–92, Jan. 1993, doi:
10.1006/jcph.1993.1010.
[51] E.-G. Talbi, Metaheuristics. Hoboken, NJ, USA: John Wiley & Sons, Inc., 2009.
[52] M. Alzaqebah, S. Jawarneh, M. Alwohaibi, M. K. Alsmadi, I. Almarashdeh, and R. Mustafa A. Mohammad, “Hybrid brain storm
optimization algorithm and late acceptance hill climbing to solve the flexible job-shop scheduling problem,” Journal of King Saud
University-Computer and Information Sciences, Sep. 2020, doi: 10.1016/j.jksuci.2020.09.004.
[53] E. Emary, H. M. Zawbaa, and A. E. Hassanien, “Binary ant lion approaches for feature selection,” Neurocomputing, vol. 213, pp.
54–65, Nov. 2016, doi: 10.1016/j.neucom.2016.03.101.
[54] M. Alwohaibi, M. Alzaqebah, N. M. Alotaibi, A. M. Alzahrani, and M. Zouch, “A hybrid multi-stage learning technique based on
brain storming optimization algorithm for breast cancer recurrence prediction,” Journal of King Saud University-Computer and
Information Sciences, May 2021, doi: 10.1016/j.jksuci.2021.05.004.
[55] “UCI machine learning repository.” http://archive.ics.uci.edu/ml/index.php (accessed Dec. 25, 2021).
[56] M. Alzaqebah et al., “Memory based cuckoo search algorithm for feature selection of gene expression dataset,” Informatics in
Medicine Unlocked, vol. 24, 2021, doi: 10.1016/j.imu.2021.100572.
[57] M. Mafarja et al., “Evolutionary population dynamics and grasshopper optimization approaches for feature selection problems,”
Knowledge-Based Systems, vol. 145, pp. 25–45, Apr. 2018, doi: 10.1016/j.knosys.2017.12.037.
[58] E. Hato, “Impact of feature selection for data classification using naive Bayes classifier,” Journal of Physics: Conference Series,
vol. 1879, no. 2, May 2021, doi: 10.1088/1742-6596/1879/2/022088.
[59] F. Han, W.-T. Chen, Q.-H. Ling, and H. Han, “Multi-objective particle swarm optimization with adaptive strategies for feature
selection,” Swarm and Evolutionary Computation, vol. 62, Apr. 2021, doi: 10.1016/j.swevo.2021.100847.
[60] T. M. T. A. Hamid, R. Sallehuddin, Z. M. Yunos, and A. Ali, “Ensemble based filter feature selection with harmonize particle
swarm optimization and support vector machine for optimal cancer classification,” Machine Learning with Applications, vol. 5,
Sep. 2021, doi: 10.1016/j.mlwa.2021.100054.

4325
BIOGRAPHIES OF AUTHORS
Mutasem Khalil Alsmadi is currently an associate professor at the Faculty of
Applied Studies and Community Service, Department of Management of Information Systems,
Imam Abdurrahman Bin Faisal University. He received his B.S. degree in Software
engineering in 2006 from Philadelphia University, Jordan, his M.Sc. degree in intelligent
systems in 2007 from University Utara Malaysia, Malaysia, and his Ph.D. in Computer Science
from The National University of Malaysia. He has published more than one hundred papers in
the image processing and algorithm optimization areas. His research interests include artificial
intelligence, pattern recognition, algorithms optimization and computer vision. He can be
contacted at email: mksalsmadi@gmail.com.
Malek Alzaqebah obtained his B.Sc. in computer science from Al-Balqa Applied
University, Jordan and his master degree specializing in information technology from
University Utara Malaysia (UUM). He did his Ph.D. in computer science at University
Kebangsaan Malaysia (UKM). Now, he is an assistant professor at the Department of
Mathematics, Faculty of Science, Imam Abdulrahman Bin Faisal University, Dammam, Saudi
Arabia. His research interest involves data mining and metaheuristic algorithms for
combinatorial optimization problems such as university timetabling, job shop scheduling,
vehicle routing and search-based software engineering. He can be contacted at email:
maafehaid@iau.edu.sa.
Sana Jawarneh obtained her BSc in computer engineering from Yarmouk
University and her PhD in computer science at University Kebangsaan Malaysia. Now, she is
an assistant professor at Department of computer science, The applied college, Imam
Abdulrahman Bin Faisal University, Dammam, Saudi Arabia. Her research interest falls under
meta-heuristic algorithms in various optimization problems. Email: sijawarneh@iau.edu.sa.
Sami Brini obtained his B.Sc. in computer science from Faculty of Science of
Monastir, Monastir University, Tunisia and his master’s degree in computer science from
University of Sousse, Tunisia. He prepared his thesis in Tunisia that covers Artificial
Intelligence and NLP. He is currently a Faculty Member at the Department of Mathematics,
Faculty of Science, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia. His
research interest involves data analysis and security, computational physics and modeling,
machine learning, computing and artificial intelligence. He can be contacted at email:
ssbrini@iau.edu.sa.
Ibrahim Abood Al-Marashdeh is currently an associate professor at the Faculty
of Applied Studies and Community Service, Department of Management of Information
Systems, Imam Abdurrahman Bin Faisal University. He received his B.Sc from University of
Jordan in 2001. His MSc degree in information and communication technology in in 2007
from University Utara Malaysia, and his PhD in Software engineering/Industrial computing
from The National University of Malaysia award in 2012. He has published number of
research papers and his research include: human computer interaction, image processing and
algorithm optimization areas and software development and usability, requirements analysis,
mobile government and mobile learning. he can be contacted at email: ibramars@gmail.com.

 ISSN: 2088-8708
4326
Khaoula Briki obtained his BSc in computer science from the Higher Institute of
Applied Sciences and Technology of Sousse (ISSATS), University of Sousse her master’s
degree specializing in Decision Making from Higher Institute of Computer Science and
Management of Kairouan (ISIGK), University of Kairouan. Now, she is a Lecturer at the
Department of Mathematics, Faculty of Science, Imam Abdulrahman Bin Faisal University,
Dammam, Saudi Arabia. She is currently Head of the Learning Outcomes Measurement Unit,
College of Science. Prior to this, she was Head of Statistical Analysis and Measurement Unit,
College of Science. Her main research interests and experiences are in diverse areas of
machine intelligence, data mining and data analysis. She can be contacted at email:
kabriki@iau.edu.sa.
Nashat Alrefai received the B.Sc. in programming and system analysis from
Jordan University of Science and Technology. He got his Master’s degree in computer
information systems from the University of Banking and Financial Sciences and he is a Ph.D.
candidate in University Technology Malaysia (UTM). He is a lecturer at the Department of
Mathematics, Faculty of Science, Imam Abdulrahman Bin Faisal University, Dammam, Saudi
Arabia, and work as researcher in basic and applied scientific research center. His research
focuses on deep learning, machine learning, artificial intelligence, bioinformatics and data
mining. Email: nalrefai@iau.edu.sa.
Fahad Ali Alghamdi is currently an assistant at the Faculty of Applied Studies
and Community Service, Department of Management of Information Systems, Imam
Abdurrahman Bin Faisal University, Saudi Arabia. He received his B.S. degree in computer
science in 2002 from Imam Abdurrahman Bin Faisal University, Saudi Arabia, his M.Sc.
degree in computer science in 2009 from University Wollongong, Australia, and his Ph.D. in
Computer Science in 2017 from Aberystwyth University in UK. His research interests include
network, applied computer software, cloud computing and online learning. He is currently the
Vice Dean of Faculty of Applied Studies and Community Service, Imam Abdurrahman Bin
Faisal University, Saudi Arabia. He can be contacted at email: faghamdi@iau.edu.sa.
Maen T. Al-Rashdan received the BEng. degree in Electronic engineering from
Yarmouk University of Jordan, in 2001, and the M.Sc. degree in Computer and
Communication Engineering from the National University of Malaysia (UKM), in 2004 and
completed his PhD in Computer and Communication Engineering, electrical, electronic and
system engineering department, UKM, Malaysia, in 2011. His research has been focused on
Smart Selection Hierarchical Mobile Internet Protocol Version Six with Load Balancing and
Mobility Anchor Points Queue Management. He is currently working as Assisstant Professor
at Jadara University. He has published several articles in top journals and international
conferences. He can be contacted at Email: dr.maen@jadara.edu.jo.

Cuckoo algorithm with great deluge local-search for feature selection problems

More Related Content

Similar to Cuckoo algorithm with great deluge local-search for feature selection problems (20)

More from IJECEIAES (20)

Recently uploaded (20)

Cuckoo algorithm with great deluge local-search for feature selection problems