IAES International Journal of Artificial Intelligence (IJ-AI)

Vol. 13, No. 2, June 2024, pp. 2395~2402

ISSN: 2252-8938, DOI: 10.11591/ijai.v13.i2.pp2395-2402  2395

Feature selection techniques for microarray dataset: a review

Avinash Nagaraja Rao1, Sitesh Kumar Sinha1, Shivamurthaiah Mallaiah2

1
Department of Computer Science and Engineering, Rabindranath Tagore University, Bhopal, India
2
Department of Computer Science, APS College of Engineering, Bengaluru, Bangalore, India

Article Info ABSTRACT

Article history: Automatic speech recognition (ASR) approach is dependent on optimal
for many researchers working on feature selection (FS) techniques,
Received Jun 30, 2023
finding an appropriate feature from the microarray dataset has turned
Revised Oct 12, 2023
into a bottleneck. Researchers often create FS approaches and
Accepted Dec 2, 2023
algorithms with the goal of improving accuracy in microarray datasets.
The main goal of this study is to present a variety of contemporary FS
Keywords: techniques, such as filter, wrapper, and embedded methods proposed for
microarray datasets to work on multi-class classification problems and
Correlation different approaches to enhance the performance of learning algorithms,
Embedded to address the imbalance issue in the data set, and to support research
Feature selection efforts on microarray dataset. This study is based on critical review
Filter questions (CRQ) constructed using feature election methods described
Microarray dataset in the review methodology and applied to a microarray dataset. We
Wrapper discussed the analysed findings and future prospects of FS strategies for
multi-class classification issues using microarray datasets, as well as
prospective ways to speed up computing environment.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Avinash Nagaraja Rao
Department of Computer Science and Engineering, Rabindranath Tagore University
Bhopal, India
Email: avi003@gmail.com

1. INTRODUCTION
Feature is an important element in the classification problem addressed in machine learning (ML)
techniques. Selecting the best feature or relevant feature will provide better accuracy in optimization algorithm
[1]. Microarray dataset contains minimum number of samples and maximum number of features. Conventionally,
most of the microarray datasets include more than 60,000 characteristics or attributes with fewer samples, not
surpassing 100 [2]. Due to this scenario important or relevant features are identified using feature selection (FS)
techniques viz., filter, wrapper, and embedded methods. This leads to significant reduction in obtaining the
optimal results with classification accuracy and computational time [3]. The primary intention of this review is:
i) to provide the various FS algorithms to find the best accuracy in microarray dataset analysis, ii) to introduce the
various ML techniques involved in the microarray dataset, iii) to examine the publication trends in the domain
related to FS using microarray dataset, iv) to consolidate the various research work in different FS techniques by
various authors and experts, and v) to provide future scope of research in real time microarray dataset.
Optimum results are obtained in the microarray dataset using appropriate FS techniques. The
conventional FS techniques like filter method, wrapper method, embedded method, and hybrid methods are
favoured for the subset selection to obtain optimal results. Figure 1 depicts the standard procedure of a FS
process. The first step is to use a reliable search strategy to create a subset of the microarray dataset. The second
stage involves evaluating the list of subgroups and contrasting the best subset with the one that came before it.

Journal homepage: http://ijai.iaescore.com

2396  ISSN: 2252-8938

If the newly updated subset is highly suggested than the previous one, then the change is left unmodified. Until
the termination condition is met, the procedure will be repeated. Then finally, the best subset score is selected
and given to the next classification technique. Selecting a feature subset is achieved using the Algorithm 1 [4].

Figure 1. General FS process

Figure 2 shows the complete work flow of filter, wrapper, and embedded model, which finds the best
subset in all the features available in the dataset. Both the wrapper and embedded methods have the stopping
condition to validate the best subset [5]. Filter method does not involve with particular learning algorithm and
validation of best subset whereas wrapper method selects the features using the learning algorithm and
validating the optimal feature subset [6]. In embedded methods combines the both filter and wrapper method.
As per the study we have represented inferences and future scope of each method in the Table 1.

Algorithm 1: Selecting feature subset

Input: large future microarray dataset
Output: Obtain the best subset
Step 1: Starts with subset generation
Step 2: Subset evaluation
Step 3: Stopping condition
Step 4: Validating the feature subset
if condition satisfies, test n number of times
Select the best feature set
else
Repeat the step2 and step3
Step 5: End process

Table 1. Inference and future scope of FS methods

Method Inference Future scope
Filter As per the review with the microarray dataset, majority It is observed that the researchers can work on research
of researchers (about 40%) used filter method to obtain problems like high computational cost and runtime
the optimal classification performance 2015. performance.
Wrapper As per the review 20% of the researchers have Multi-objective optimization is to be experimented with
experimented with the wrapper methods using well the various microarray datasets and results should be
known population methods to achieve better compared with other algorithms.
classification accuracy.
Embedded As per the review 40% of the researchers have used the
embedded FS technique in their microarray dataset

Int J Artif Intell, Vol. 13, No. 2, June 2024: 2395-2402

Int J Artif Intell ISSN: 2252-8938  2397

Figure 2. Basic flow of filter, wrapper, embedded methods

2. LITERATURE REVIEW
This section dives into two key aspects of analyzing microarray data: FS techniques and the overall
review technology roadmap. We'll explore the fundamental methods for selecting the most informative genes
from these VAST datasets, followed by a roadmap outlining the steps involved in effectively reviewing and
analyzing this technology.

2.1. Filter method

The filter method checks whether the particular feature is relevant or not in the microarray dataset
apart from calculating the error rate. Initially, feature/relevance score is calculated and then it is compared with
the remaining columns in the dataset, then it finalizes the low feature/relevance scoring subset to be ignored
form the microarray dataset. The optimal subset features are fed into the classification technique for the next
process. The filter methods can be classified into two types, namely univariate and multivariate. The univariate
evaluates the features separately and the multivariate filters evaluate features concerning the relationship
between the two features [7]. Feature scoring is calculated using the following methods like pearson
correlation, chi-squared test, entropy, Fisher’s score, Spearman correlation, Kendall correlation, analysis of
variance (ANOVA), count based, linear discriminant analysis (LDA), relief, information gains fast correlation-
based filter and maximal relevance and minimal redundancy (mRMR) [8].

2.2. Wrapper method

In the wrapper method, subset features in the dataset are evaluated independently. The evaluation
criteria of the specific subset are achieved by using the learning algorithm(s). Due to the exponential growth
of the subset, the number of features will increase in this scenario. The learning algorithms are used to find the
best fitness of the subset [9]. The wrapper model algorithm gives more understanding of the selection of the
best subset in the features. In the wrapper model, the search methods are classified into two major categories,
namely sequential and randomized. The sequential search methods consist following algorithms like sequential
forward selection (SFS), sequential backward selection (SBS), plus q take away r and beam search. The second
method is a population-based algorithm or sequential search [10] that finds better optimum fitness of the subset,
using techniques such as genetic algorithm, particle swarm algorithm, advanced binary ant colony

Feature selection techniques for microarray dataset ... (Avinash Nagaraja)

2398  ISSN: 2252-8938

optimization, harmony [11], differential evolution, whale optimization, artificial bee colony, and bacterial
colony optimization [12].
2.3. Embedded method
The embedded FS technique is a classifier dependent FS method [13]. The learning algorithm is
playing a vital role in the embedded method. The researchers prefer the embedded method commonly due to
the low computational cost. Irrelevant features are removed using widely used techniques like weight vector
of support vector machine (SVM), decision tree, weighted Naive Bayes (NB). The following embedded
methods, namely first-order inductive learner (FOIL) feature subset selection algorithm, probably
approximately correct (PAC) Bayes, kernel-penalized support vector machine (KP-SVM), least absolute
shrinkage and selection operator (LASSO).
The committed writing uncovers an assortment of approaches managing with critical resource
questions (CRQ) issues within this section details about review methodology employed throughout this paper.
As per the intension framed for this study along with CRQ is framed and the road map of the review
methodology is represented in Table 2. In the initial stage we have framed the CRQ based FS techniques used
in microarray dataset. While in the second stage we have selected 41 manuscripts related to the topic in recent
years. In the next stage FS techniques used in these articles are analyzed. Then we have focused on the
classification accuracy using various ML techniques. The final stage of the review methodology is to provide
the future directions to the researchers, from the insights. The review is based on the key questions which help
to form a study of this paper is represented in Table 2.

Table 2. Research questions

ID Question Motivation
CRQ1 What are the recent publications that interpret the outcome To examine the publication trends in the domain related to
and feature scopes to the benefit of researchers? FS using microarray dataset.
CRQ2 What are the different FS techniques used in microarray To provide the various FS algorithms to find the best
dataset? accuracy in microarray dataset analysis
CRQ3 What are the common microarray datasets used in recent To provide the frequently used microarray dataset in recent
years? years.
CRQ4 What tools are used to implement the FS techniques when we To explore the tools used by the researchers in high
select high dimensional dataset? dimensional dataset.
CRQ5 What is the future scope of research in real time microarray To provide the future scope to the researchers based on this
dataset? study.

2.4. Literature resources

This section is to identify the candidate papers which can helps to answer the CRQs represented in
the Table 2. Research articles are collected from different online sources like Scopus, association for computing
machinery (ACM) digital library, and PubMed. The selection of the research paper is done using search
selection using keywords, tittle, and abstract. Another criterion for selection is we have year constrains starts
from 2014 to 2022.

2.5. Search process and study selection

The automatic search process is done with different online sources with the two criteria explained.
The potential research papers are identified and collected for the study. The frequent checks are done with
online sources and confirmed that there are relevant contents are collected for our study. Based on those content
we have formed two criteria named inclusion criteria and exclusion criteria for our study.

2.6. Inclusion criteria

Inclusion criteria are characteristics that are to be included in the study. In our study we have included
following characterstics. i) research articles which address the microarray dataset (mandatory), ii) research
articles using FS techniques, iii) research articles performing comparison of various FS techniques on different
microarray dataset, and iv) research articles which are published recent years.

2.7. Exclusion criteria

Exclusion criteria comprise characteristics used to identify potential research participants who should
not be included in a study. In our study following are comes under exclusion criteria: i) research articles which
are not written in English language and ii) research articles which contain less FS techniques. The mandatory
criterion helps us to select the papers and if any one of the exclusion criteria is satisfied means we have rejected
the papers for our study. Initially the tittle, keywords, and abstracts are examined and if it is required further
study is done on the full article.

Int J Artif Intell, Vol. 13, No. 2, June 2024: 2395-2402

Int J Artif Intell ISSN: 2252-8938  2399

2.8. Data extraction

After identifying the relevant research articles which satisfied our CRQs information’s are extracted
from the selected papers. We have created a checklist depends on our CRQs represented in Table 3. The aim
of the check list is to extract sufficient information on the CRQs which is maintained in the separate excel
sheet. We have extracted major contents to our study in this section. The second phase is cross validation
between the authors. We are acknowledged that all the selected papers are not met the five CRQs.

2.9. Recent micro arraydatasets

Datasets plays a vital role in any ML applications [14]. The curse of dimensionality is the primary
research issue, faced by researchers while dealing with a microarray dataset [15]. Due to the high number of
samples, it is hard to find the optimal results with a high dimensional dataset. Generally, microarray datasets
are grouped into two different categories named binary classification and multiclass classification. Many
researchers have experimented in binary class classification [16].

2.10 Challenges
The following obersrvation are the challenges faced during the exection features and dataset: i) the
computational time is high due to a greater number of features in the gene expression with the noise, and ii)
the unbalanced dataset will affect the training and test dataset so that the accuracy will be another issue.
Table 4 (see an Appendix) explains review methodology, for this different journals were refered from each
different methods were proposed for FS and accuracy is measured from each method and feature
enhanchements that can be done from each were dissussed.

Table 3. Check list form

CRQ Validation
1 What are the recent publications that interpret the outcome and feature scopes to the benefit of researchers?
‒ Year
‒ Source name
‒ Feature scope
2 What are the different FS techniques used in microarray dataset?
‒ Filter
‒ Wrapper
‒ Embedded
‒ Hybrid
3 What are the common microarray datasets used in recent years?
‒ Dataset name
‒ Binary classification 1 0
‒ Multi class classification
4 What tools are used to implement the FS techniques when we select high dimensional dataset?
‒ Name of the tools used to perform FS
5 What is the future scope of research in real time microarray dataset?
‒ Future scope to implement in real time dataset

3. RESEARCH PERSPECTIVE AFTER LITREATURE REVIEW

3.1. Stability
When developing FS approach for high-dimensional datasets, stability is of paramount importance. If
the FS method yields the same result independent of changes to the input data, it is considered resilient [26]. If
the stability issue of the FS method is ignored, erroneous inference and inaccurate results may result. One of
the most glaring causes of unpredictability is the neglect of features that were selected because they were
aligned with the dependent variables. Stability facilitates the trade-off between bias and volatility in the
classification error rate [27]. The dimensionality of the dataset, the number of features selected, the size of the
sample, and the unpredictability of the data all have a role in the FS algorithm's stability [28].

3.2. Choosing appropriate measures of fitness

Wrapper FS techniques seek to optimise a feature subset by maximising an objective function. The
classification problem influences the development of the FS goal function. At first, an objective function was
developed to maximise classification success while minimising the number of features. Feature counts and
classification accuracy are merged into a single fitness function in a single-objective technique [29].

3.3. Selection of classifiers

Various classifiers have been used to address FS issues, including k-nearest neighbour (KNN), NB,
SVM, random forest (RF), and artificial neural networks (ANN). The most crucial step in optimising results
from a high-dimensional dataset is selecting a classifier to use. The KNN was found to be the most popular
classifier in the reviewed literature, and in the classification process, SVM is essential [30].
Feature selection techniques for microarray dataset ... (Avinash Nagaraja)
2400  ISSN: 2252-8938

3.4. Reduced size of sample

Numerous studies have sought to address the widespread problem of limited data from individual
microarray experiments. One primary worry is that learning methods suffer significantly when working with little
data. The right validation strategy for calculating the misclassification rate is essential for fixing this problem [31].

3.5. Disparity in class

An unfavourable learning environment is created when one class dominates the dataset at the expense
of the others. Unbalanced microarray datasets [32] are common, and multiclass microarray datasets are a prime
example. When the test set's imbalance is more pronounced than in the training sets, solving this issue becomes
difficult. To get around this issue, several researchers use preprocessing techniques like under sampling and
oversampling. Ensemble classifiers have recently been proposed as a way to address class imbalance [33].

3.6. Outliers
One of the most important yet under-discussed aspects of microarray data is the identification of
outliers. Polluted database samples are known as outliers, and they occur when instruments or humans make
mistakes during data collection or analysis [34]. The learning process is hindered by outliers because they
prevent the useful genes from being chosen.

4. DISCUSSION AND FUTURE DIRECTIONS

The examination of microarray data provides important clues that facilitate the identification of new
diseases. High complexity of gene expression data and small sample size make categorization a difficult
process. Therefore, FS strategy is the most workable answer to these issues. This research meticulously
synthesises the approaches, techniques, datasets, and future possibilities of the microarray dataset during the
last three years. Reviewing the literature critically, this article examines a wide range of topics, including
multiclass classification, methods for enhancing learning algorithm performance, resolving the issue of data
set imbalance, and verifying the results of previous work using microarray datasets. This work provides a
comprehensive examination of FS procedures in the context of search strategies, offering insights into the
current techniques used for FS on microarray datasets. This paper presents a complete literature evaluation of
five distinct kinds of FS strategies, including filter, wrapper, and embedding methods. This paper examines the
challenges and research considerations that arise during the development of a FS method.

APPENDIX

Table 4. Review methodology

S.no Ref Proposed method Evaluation Findings or future enhancements
measures
1 [17] Improved version of salp swarm algorithm (SSA) Accuracy and The converging ability is not investigated. In
with transfer function and cross over scheme to fitness value addition, S-Shaped and V-shaped transfer
overcome the issues of FS. functions are to be explored in future
enhancement.
2 [18] Hybrid variant of Harris Hawk optimization Accuracy and High dimensional datasets are not
(HHO) based on bitwise and simulated annealing fitness value investigated.
(SA) to solve the FS problem.
3 [19] Improved SSA with inertia weight parameter to Classification The converging ability of the model is not
enhance both exploration and exploitation accuracy and tested.
capabilities. feature reduction.
4 [20] Binary equilibrium optimization combined with Accuracy and High dimensional datasets are not
SA for FS problems. fitness value investigated.
5 [21] A novel single objective optimization based on Accuracy and The medical data is not experimented in the
chaotic theory for FS problems. fitness value proposed system.
6 [21] Opposition-based self-adaptive cohort intelligence Accuracy and High dimensional datasets are not
(OSCI), optimal features have been selected. fitness value investigated.
7 [22] Dynamic version of SSA to solve the local optima Accuracy and The updating rules can be modified based on
problem to balance both exploration and exploitation. fitness value the problem statement.
8 [23] Hybrid version of SSA and simulated annealing to Accuracy and The model has been implemented for multi
improve the exploitation capability. number of features objective optimization.
are tested
9 [24] Binary version of SSA with three updating rules Accuracy, fitness The exploration search space is improved.
to solve the FS problems. value, and Furthermore, updating rules can be modified
converging ability. based on the problem statement.
10 [25] Binary version of SSA with random weight Classification The investigation is performed on binary
network (RWN) to solve the FS problem. accuracy classification problems. The exploration and
exploitation are not considered for
evaluating the model.

Int J Artif Intell, Vol. 13, No. 2, June 2024: 2395-2402

Int J Artif Intell ISSN: 2252-8938  2401

REFERENCES
[1] M. A. Hambali, T. O. Oladele, and K. S. Adewole, “Microarray cancer feature selection: review, challenges and research directions,”
International Journal of Cognitive Computing in Engineering, vol. 1, pp. 78–97, 2020, doi: 10.1016/j.ijcce.2020.11.001.
[2] V. Bolón-Canedo and B. Remeseiro, “Feature selection in image analysis: a survey,” Artificial Intelligence Review, vol. 53, no. 4,
pp. 2905–2931, 2020, doi: 10.1007/s10462-019-09750-3.
[3] R. C. Chen, C. Dewi, S. W. Huang, and R. E. Caraka, “Selecting critical features for data classification based on machine learning
methods,” Journal of Big Data, vol. 7, no. 1, pp. 1–26, 2020, doi: 10.1186/s40537-020-00327-4.
[4] B. Remeseiro and V. Bolon-Canedo, “A review of feature selection methods in medical applications,” Computers in Biology and
Medicine, vol. 112, pp. 1–9, 2019, doi: 10.1016/j.compbiomed.2019.103375.
[5] S. Shadravan, H. R. Naji, and V. K. Bardsiri, “The sailfish optimizer: a novel nature-inspired metaheuristic algorithm for solving
constrained engineering optimization problems,” Engineering Applications of Artificial Intelligence, vol. 80, pp. 20–34, 2019, doi:
10.1016/j.engappai.2019.01.001.
[6] K. Tadist, S. Najah, N. S. Nikolov, F. Mrabti, and A. Zahi, “Feature selection methods and genomic big data: a systematic review,”
Journal of Big Data, vol. 6, no. 1, pp. 1–24, 2019, doi: 10.1186/s40537-019-0241-0.
[7] T. Saw and P. Hnin, “Swarm intelligence based feature selection for high dimensional classification: a literature survey,”
International Journal of Computer (IJC), vol. 33, no. 1, pp. 69–83, 2019.
[8] Y. Saeys, I. Inza, and P. Larrañaga, “A review of feature selection techniques in bioinformatics,” Bioinformatics, vol. 23, no. 19,
pp. 2507–2517, 2007, doi: 10.1093/bioinformatics/btm344.
[9] A. Mangal and E. A. Holm, “A comparative study of feature selection methods for stress hotspot classification in materials,”
Integrating Materials and Manufacturing Innovation, vol. 7, no. 3, pp. 87–95, 2018, doi: 10.1007/s40192-018-0109-8.
[10] R. R. Rani and D. Ramyachitra, “Microarray cancer gene feature selection using spider monkey optimization algorithm and cancer
classification using SVM,” Procedia Computer Science, vol. 143, pp. 108–116, 2018, doi: 10.1016/j.procs.2018.10.358.
[11] Y. Prasad, K. K. Biswas, and M. Hanmandlu, “A recursive PSO scheme for gene selection in microarray data,” Applied Soft
Computing Journal, vol. 71, pp. 213–225, 2018, doi: 10.1016/j.asoc.2018.06.019.
[12] B. Sahu, S. Dehuri, and A. K. Jagadev, “Feature selection model based on clustering and ranking in pipeline for microarray data,”
Informatics in Medicine Unlocked, vol. 9, pp. 107–122, 2017, doi: 10.1016/j.imu.2017.07.004.
[13] M. K. Ebrahimpour, H. Nezamabadi-pour, and M. Eftekhari, “CCFS: a cooperating coevolution technique for large scale feature selection
on microarray datasets,” Computational Biology and Chemistry, vol. 73, pp. 171–178, 2018, doi: 10.1016/j.compbiolchem.2018.02.006.
[14] C. Arunkumar and S. Ramakrishnan, “Attribute selection using fuzzy rough set based customized similarity measure for lung cancer
microarray gene expression data,” Future Computing and Informatics Journal, no. 3, pp. 131–142, 2018.
[15] H. Dong, T. Li, R. Ding, and J. Sun, “A novel hybrid genetic algorithm with granular information for feature selection and
optimization,” Applied Soft Computing Journal, vol. 65, pp. 33–46, 2018, doi: 10.1016/j.asoc.2017.12.048.
[16] S. Maldonado, R. Weber, and F. Famili, “Feature selection for high-dimensional class-imbalanced data sets using support vector
machines,” Information Sciences, vol. 286, pp. 228–246, 2014, doi: 10.1016/j.ins.2014.07.015.
[17] R. J. Urbanowicz, M. Meeker, W. La Cava, R. S. Olson, and J. H. Moore, “Relief-based feature selection: introduction and review,”
Journal of Biomedical Informatics, vol. 85, pp. 189–203, 2018, doi: 10.1016/j.jbi.2018.07.014.
[18] E. H. Houssein, M. E. Hosney, M. Elhoseny, D. Oliva, W. M. Mohamed, and M. Hassaballah, “Hybrid Harris hawks optimization
with cuckoo search for drug design and discovery in chemoinformatics,” Scientific Reports, vol. 10, no. 1, pp. 1–22, 2020, doi:
10.1038/s41598-020-71502-z.
[19] A. E. Hegazy, M. A. Makhlouf, and G. S. El-Tawel, “Improved salp swarm algorithm for feature selection,” Journal of King Saud
University - Computer and Information Sciences, vol. 32, no. 3, pp. 335–344, 2020, doi: 10.1016/j.jksuci.2018.06.003.
[20] Y. Gao, Y. Zhou, and Q. Luo, “An efficient binary equilibrium optimizer algorithm for feature selection,” IEEE Access, vol. 8, pp.
140936–140963, 2020, doi: 10.1109/ACCESS.2020.3013617.
[21] A. E. Hegazy, M. A. Makhlouf, and G. S. El-Tawel, “Feature selection using chaotic salp swarm algorithm for data classification,”
Arabian Journal for Science and Engineering, vol. 44, no. 4, pp. 3801–3816, 2019, doi: 10.1007/s13369-018-3680-6.
[22] R. Storn and K. Price, “Differential evolution – a simple and efficient heuristic for global optimization over continuous spaces,”
Journal of Global Optimization, no. 11, pp. 341–359, 1997.
[23] S. Mirjalili and A. Lewis, “The whale optimization algorithm,” Advances in Engineering Software, vol. 95, pp. 51–67, 2016, doi:
10.1016/j.advengsoft.2016.01.008.
[24] W. Gao, S. Liu, and L. Huang, “A global best artificial bee colony algorithm for global optimization,” Journal of Computational
and Applied Mathematics, vol. 236, no. 11, pp. 2741–2753, 2012, doi: 10.1016/j.cam.2012.01.013.
[25] S. H. Bouazza, K. Auhmani, A. Zeroual, and N. Hamdi, “Selecting significant marker genes from microarray data by filter approach
for cancer diagnosis,” Procedia Computer Science, vol. 127, pp. 300–309, 2018, doi: 10.1016/j.procs.2018.01.126.
[26] A. K. Shukla, P. Singh, and M. Vardhan, “A hybrid gene selection method for microarray recognition,” Biocybernetics and
Biomedical Engineering, vol. 38, no. 4, pp. 975–991, 2018, doi: 10.1016/j.bbe.2018.08.004.
[27] M. J. Rani and D. Devaraj, “Two-stage hybrid gene selection using mutual information and genetic algorithm for cancer data
classification,” Journal of Medical Systems, vol. 43, no. 8, 2019, doi: 10.1007/s10916-019-1372-8.
[28] S. Maldonado and J. López, “Dealing with high-dimensional class-imbalanced datasets: embedded feature selection for SVM
classification,” Applied Soft Computing Journal, vol. 67, pp. 94–105, 2018, doi: 10.1016/j.asoc.2018.02.051.
[29] J. R. Anaraki and H. Usefi, “A comparative study of feature selection methods on genomic datasets,” in Proceedings - IEEE
Symposium on Computer-Based Medical Systems, 2019, pp. 471–476, doi: 10.1109/CBMS.2019.00097.
[30] R. Aziz, C. K. Verma, and N. Srivastava, “Dimension reduction methods for microarray data: a review,” AIMS Bioengineering, vol.
4, no. 1, pp. 179–197, 2017, doi: 10.3934/bioeng.2017.1.179.
[31] K. Balakrishnan, R. Dhanalakshmi, and U. M. Khaire, “Improved salp swarm algorithm based on the levy flight for feature
selection,” Journal of Supercomputing, vol. 77, no. 11, pp. 12399–12419, 2021, doi: 10.1007/s11227-021-03773-w.
[32] M. Liu, X. Yao, and Y. Li, “Hybrid whale optimization algorithm enhanced with Lévy flight and differential evolution for job shop
scheduling problems,” Applied Soft Computing Journal, vol. 87, 2020, doi: 10.1016/j.asoc.2019.105954.
[33] B. Seijo-Pardo, I. Porto-Díaz, V. Bolón-Canedo, and A. Alonso-Betanzos, “Ensemble feature selection: homogeneous and
heterogeneous approaches,” Knowledge-Based Systems, vol. 118, pp. 124–139, 2017, doi: 10.1016/j.knosys.2016.11.017.
[34] F. Yang and K. Z. Mao, “Robust feature selection for microarray data based on multicriterion fusion,” IEEE/ACM Transactions on
Computational Biology and Bioinformatics, vol. 8, no. 4, pp. 1080–1092, 2011, doi: 10.1109/TCBB.2010.103.

Feature selection techniques for microarray dataset ... (Avinash Nagaraja)

2402  ISSN: 2252-8938

BIIOGRAPHIES OF AUTHORS

Mr. Avinash Nagaraja Rao is a research scholar in the Department of

Computer Science and Engineering, RNTU, Bhopal. He has obtained his M.Tech. degree
in Computer Science and Engineering from VTU. He has 13 years of teaching experience
and published more than 10 research papers in reputed international journals. His main
research work focuses on machine learning, deep learning, network security, image
processing, and data mining. He can be contacted at email: avi003@gmail.com.

Dr. Sitesh Kumar Sinha completed his Ph.D. from degree from BRAB MIT,
Muzaffarpur, Bihar. Currently, he working as registrar (administration) Dr C. V. Raman
University Vaishali Bihar. He is completed government funding project during his Ph.D.
work related on computer network. He is published more than 40 research paper in various
international and national journals. His main research work focuses on network security,
image processing, and software engineering. He can be contacted at email:
siteshkumarsinha@gmail.com.

Dr. Shivamurthaiah M completed his Ph.D. from RNTU and currently

working as professor in Department of CSE at Dr. S. M. College of Engineering, Bengaluru.
He has 17 years of teaching and 9 years of research experience and published more than 30
research papers in reputed international journals. He has published 9 patents in that 2 patent
are granted. His main research work focuses on cryptography algorithms, network security,
image processing and software engineering data mining, IoT, and computational
intelligence-based education. He can be contacted at email: shivamurthaiah@gmail.com.

Int J Artif Intell, Vol. 13, No. 2, June 2024: 2395-2402