IAES International Journal of Artificial Intelligence (IJ-AI)

Vol. 14, No. 1, February 2025, pp. 466~472

ISSN: 2252-8938, DOI: 10.11591/ijai.v14.i1.pp466-472  466

Revolutionizing cancer classification: the snr-ogscc method for

improved gene selection and clustering

Sara Haddou Bouazza, Jihad Haddou Bouazza

LAMIGEP, EMSI, Marrakech, Morocco

Article Info ABSTRACT

Article history: This study presents the signal-to-noise ratio optimized gene selection and
clustering for cancer classification (SNR-OGSCC) methodology, aimed at
Received Sep 9, 2024 enhancing classification accuracy while reducing the dimensionality of gene
Revised Oct 19, 2024 expression data across various cancer types. Implemented on a standard
Accepted Oct 23, 2024 computational setup, the SNR-OGSCC method combines advanced filtering,
clustering, and machine learning techniques, demonstrating significant
improvements in classification accuracy on seven cancer datasets: leukemia,
Keywords: colon cancer, prostate cancer, lung cancer, lymphoma, central nervous system
(CNS) tumors, and ovarian cancer. Notably, our approach achieved perfect
Artificial intelligence accuracies of 100% for leukemia, lung cancer, and ovarian cancer, with high
Cancer classification accuracies of 98.4% for colon cancer, 99.1% for prostate cancer, 98.3% for
Computer science lymphoma, and 99.7% for CNS tumors, while requiring as few as 4–5 genes
Feature selection for effective classification. These findings highlight the efficiency and
Machine learning robustness of the SNR-OGSCC methodology, suggesting its potential to
identify meaningful biomarkers and improve personalized cancer treatment
strategies. Further validation with larger datasets and biological experiments
is essential to confirm its applicability in clinical settings.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Sara Haddou Bouazza
LAMIGEP, EMSI
Marrakech, Morocco
Email: [email protected]

1. INTRODUCTION
Cancer remains one of the leading causes of morbidity and mortality worldwide, necessitating the
urgent need for advanced diagnostic tools and methodologies. Traditional diagnostic methods often rely on
histopathological examination and imaging techniques, which can be time-consuming and subjective. With the
rise of genomic technologies [1], gene expression profiling has emerged as a promising avenue for enhancing
cancer diagnosis and treatment strategies [2], [3]. This study focuses on optimizing gene selection and
clustering methodologies to improve classification accuracy in cancer detection, specifically through the
introduction of the signal-to-noise ratio optimized gene selection and clustering for cancer classification
(SNR-OGSCC) approach.
Despite significant advancements in machine learning and bioinformatics, existing methods for gene
selection and classification often face challenges related to high dimensionality and noise in gene expression
data [4]. Many traditional approaches, such as the signal-to-noise ratio (SNR) method, have been employed to
filter relevant genes [5], yet they may not adequately account for the complexities inherent in cancer datasets.
Previous studies have demonstrated the effectiveness of various classification techniques [6], however, there
remains a gap in methodologies that integrate advanced filtering, clustering, and machine learning to enhance
both accuracy and interpretability.

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

Int J Artif Intell ISSN: 2252-8938  467

Current literature reveals several methodologies with varying success rates. For example, research has
shown that artificial neural networks and genetic algorithms can achieve high classification accuracies in
leukemia cancer datasets, with reported accuracies of up to 98.5% [7]. Similarly, studies on colon and prostate
cancer have employed techniques such as random forest and deep learning models, achieving accuracies of
95.16% [8] and 97.19% [9], respectively. While these results are promising, they often come at the cost of
requiring extensive computational resources and large gene sets [10]. Our research aims to bridge these gaps
by developing a methodology that not only improves classification accuracy but also reduces the number of
genes needed for effective cancer diagnosis.
This study addresses a critical need in cancer diagnostics by presenting the SNR-OGSCC
methodology. Our approach not only enhances classification accuracy but also effectively reduces the
dimensionality of gene expression data. The subsequent sections of this paper will provide a detailed
methodology, present comprehensive results across various cancer datasets, and engage in a discussion that
highlights the implications of our findings. By applying this methodology, we aim to provide compelling
evidence that SNR-OGSCC serves as a powerful tool for researchers and clinicians, ultimately paving the way
for more precise and efficient cancer diagnostics and personalized treatment strategies.

2. METHOD
In this study, we propose an innovative methodology called SNR-OGSCC that tackles the complexity
and high dimensionality of cancer classification using gene expression data. Our approach integrates advanced
filtering, clustering, and classification techniques, with an emphasis on improving classification accuracy as
the evaluation metric. By focusing on the identification of the most relevant genes, the SNR-OGSCC
methodology aims to enhance the overall classification performance [11]–[13] across multiple types of cancer,
leading to more precise predictions and better treatment strategies.

2.1. Gene selection and filtering process

The gene selection process forms the foundation of our methodology, as it directly influences the
performance of the classification models [14], [15]. We employ the SNR as the primary filtering technique to
identify the most informative genes from each dataset. SNR compares the mean difference in expression levels
between cancerous and normal samples to the variation within each group. The formula used is as (1) [16]:
̅̅̅̅̅
𝑥1𝑗 − ̅̅̅̅̅
𝑥2𝑗
𝑃(𝑥) = (1)
𝑠1𝑗 + 𝑠2𝑗

Where ̅̅̅̅
𝑥𝑦𝑗 is the mean of attribute j and syj is its standard deviation for classes y=1, 2. We select the top
15% of genes based on their SNR scores, which ensures that only the most significant features, capable of
maximizing the separation between cancerous and normal samples, are retained for further analysis.

2.2. Clustering for dimensionality reduction

Clustering is a key component in our methodology, designed to group similar genes and further reduce
the dimensionality of the dataset [17], [18]. This step is crucial in enhancing classification accuracy by focusing
on the most relevant features while minimizing noise. We utilize k-means as the primary clustering method
[19], combined with the density-based spatial clustering of applications with noise (DBSCAN) for improved
robustness [20]. To determine the optimal number of clusters for k-means, we use the elbow method [21]
alongside the silhouette score, which assesses cluster compactness and separation. Post-k-means, DBSCAN is
applied to refine the clustering process by identifying dense regions within the dataset.

2.3. Representative gene selection from clusters

Once the clusters are established, we select representative genes from each cluster based on their SNR
scores. In the case of k-means clusters, the gene with the highest SNR score is chosen, whereas for DBSCAN,
we select the most central gene in relation to the cluster centroid. This process ensures that the final set of genes
is not only biologically significant but also optimal for the subsequent classification steps, preserving the
integrity of the data.

2.4. Classification methods

For the classification of the gene sets, we apply three powerful machine learning algorithms: k-nearest
neighbors (KNN), support vector machine (SVM), and linear discriminant analysis (LDA). Each classifier is
tailored to handle high-dimensional gene expression data, leveraging different mathematical approaches. KNN
uses Euclidean distance to classify samples based on the nearest neighbors [22], SVM employs a radial basis
function (RBF) kernel to capture non-linear relationships [23], and LDA focuses on maximizing class
Revolutionizing cancer classification: the snr-ogscc method for improved gene … (Sara Haddou Bouazza)
468  ISSN: 2252-8938

separability by modeling normally distributed data [24]. This multi-classifier strategy ensures robust
performance across various datasets.

2.5. Model evaluation

We rigorously evaluate the performance of the classifiers using stratified k-fold cross-validation [25].
This approach ensures that the class distributions in the training and testing sets are preserved across all folds,
allowing for consistent and fair assessments of classification accuracy. By using accuracy as the sole evaluation
metric, we maintain a clear and objective measure of how effectively the models distinguish between cancerous
and non-cancerous samples. The classification accuracy is calculated as (2) [26]:

(TP + TN )
Accuracy = 100 ⁄(T + T + F + F ) (2)
N N N P

True positive (TP): positive samples correctly recognized. True negative (TN): negative samples correctly
recognized. False positive (FP): negative samples wrongly identified as positive. False negative (FN): positive
samples wrongly identified as negative.

2.6. Parameters for cancer datasets

Our SNR-OGSCC methodology is applied to multiple cancer datasets, such as leukemia [27], colon
cancer [28], prostate cancer [29], lung cancer [30], lymphoma [31], central nervous system (CNS) tumors [32],
and ovarian cancer [33] each with different numbers of genes and samples. For each dataset, we use the same
filtering and clustering techniques, with consistent classifier parameters Table 1. This uniform approach ensures
comparability and reproducibility across the different types of cancer data, as shown in the parameters table.

Table 1. Proposed parameters for cancer datasets

Cancer type Number of genes Number of samples K (k-means) Classifier parameters
Leukemia 7,129 72 (47 ALL, 25 AML) 3 KNN: k=3;
SVM: C=10, γ=0.01;
LDA: ncomponents=1;
Colon cancer 6,500 62 (22 normal, 40 with cancer) 4 KNN: k=5;
SVM: C=5, γ=0.01;
LDA: ncomponents=1;
Prostate cancer 12,600 102 (52 normal, 50 with cancer) 5 KNN: k=5;
SVM: C=15, γ=0.1;
LDA: ncomponents=1;
Lung cancer 12,533 181 (31 MPM, 150 ADCA) 4 KNN: k=5;
SVM: C=20, γ=0.05;
LDA: ncomponents=1;
Lymphoma 7,070 77 (58 DLBCL, 19 FL) 3 KNN: k=5;
SVM: C=10, γ=0.01;
LDA: ncomponents=1;
Cns tumors 7,129 60 (39 survivors, 21 deceased) 4 KNN: k=3;
SVM: C=5, γ=0.1;
LDA: ncomponents=1;
Ovarian cancer 15,154 253 (91 normal, 162 with cancer) 5 KNN: k=5;
SVM: C=15, γ=0.05;
LDA: ncomponents=1;

3. RESULTS AND DISCUSSION

This section presents the findings of the SNR-OGSCC methodology. We discuss the experimental
setup, results from the new filtering approach, interpretation of these results, limitations of the study, and
conclude with the implications of our findings.

3.1. Experimental setup

The experiments were conducted on a standard laptop with an Intel® Core™ i5 CPU M 250 @ 2.4
GHz dual-core processor, 4 GB of RAM, running Windows 10 (64-bit). The MATLAB R2023a software was
utilized to implement the SNR-OGSCC methodology, perform data analysis, and carry out classification tasks.
Despite limited computational power, the system successfully processed the datasets, which ranged in size
from 6,500 to over 15,000 genes and included sample sizes between 60 and 253. This setup allowed the study
to balance computational efficiency with classification accuracy, proving that even with basic hardware; the
proposed method could be implemented effectively.

Int J Artif Intell, Vol. 14, No. 1, February 2025: 466-472

Int J Artif Intell ISSN: 2252-8938  469

3.2. Results corresponding to the new filtering approach

This study evaluated the performance of the SNR-OGSCC across seven cancer datasets. The method
demonstrated substantial improvements in classification accuracy while significantly reducing the number of
genes used for each cancer type. In particular, SNR-OGSCC consistently outperformed the baseline SNR
method and the SNR+ clustering approach by optimizing gene selection through advanced clustering and
filtering techniques. Table 2 summarizes the classification accuracies and the number of genes selected by each
method (SNR, SNR+ clustering, and SNR-OGSCC) for the various cancer datasets.
The SNR-OGSCC method consistently reduced the number of genes needed for accurate
classification while maintaining or improving classification accuracy across all datasets. This demonstrates the
strength of integrating advanced clustering techniques, which significantly contribute to dimensionality
reduction and selection of the most biologically relevant genes. By refining the gene selection process and
optimizing classifiers, the method balances efficiency and performance, making it suitable for practical
applications in gene expression analysis for cancer diagnosis.

Table 2. Classification accuracies and gene selection for various cancer datasets
KNN SVM LDA
Dataset Feature selection method
Acc Nbr genes Acc Nbr genes Acc Nbr genes
Leukemia SNR 97.05 13 97.05 26 91.1 19
SNR+ clustering 100 6 100 7 100 15
SNR-OGSCC 100 4 100 5 100 4
Colon cancer SNR 92.8 5 85.7 29 92.8 2
SNR+ clustering 96 6 91.1 4 94 8
SNR-OGSCC 98.4 5 97.2 7 98.4 5
Prostate cancer SNR 90 22 92 8 92 4
SNR+ clustering 98,2 2 92 3 92 2
SNR-OGSCC 99.1 11 98.2 7 98.2 7
Lung cancer SNR 97.3 6 97.3 33 97.3 64
SNR+ clustering 100 13 99.3 10 99.3 14
SNR-OGSCC 100 5 100 7 99.3 7
Lymphoma cancer SNR 95.6 4 95.6 32 95.6 24
SNR+ clustering 97 3 97 10 97 12
SNR-OGSCC 98.3 5 98.3 9 97 6
CNS tumors SNR 76.7 6 65.1 21 69.7 28
SNR+ clustering 98.6 7 92.3 13 84 18
SNR-OGSCC 99.7 4 98.6 12 99.7 9
Ovarian cancer SNR 97.5 30 97.5 39 96.8 37
SNR+ clustering 99.3 5 100 5 97,5 11
SNR-OGSCC 100 4 100 4 99.3 9

3.3. Interpretation of results

The SNR-OGSCC method proved highly effective in improving classification accuracy across a
diverse set of cancer datasets while significantly reducing the number of genes selected. This reduction in the
number of genes directly impacts the interpretability of results and the computational load required for analysis.
For example, in the leukemia dataset, only 4-5 genes were needed to achieve 100% accuracy, highlighting the
method’s capacity to retain only the most relevant features. This improved accuracy across multiple classifiers,
including KNN, SVM, and LDA, reinforces the robustness of the proposed methodology. Moreover, the
consistent performance of SNR-OGSCC across different cancer types demonstrates its flexibility and
adaptability, making it a valuable tool for both researchers and clinicians.
The use of clustering to enhance gene selection efficiency is particularly important in complex datasets
where high dimensionality can obscure meaningful patterns. The SNR-OGSCC method not only improves
classification results but also facilitates the identification of potential biomarkers, which could have significant
implications for personalized medicine. The high classification accuracy, even with a reduced set of genes,
suggests that the method can help isolate biologically significant markers with greater precision.

3.4. Limitations
Despite the promising results, the SNR-OGSCC method has several limitations. First, the relatively
small sample sizes in certain datasets, such as CNS tumors, could limit the generalizability of the findings.
Applying the method to larger datasets would provide more robust evidence of its efficacy across a broader range
of cancer types. Additionally, while the method performed well on standard hardware, larger and more complex
datasets may require higher computational power, which could challenge the scalability of the approach.
Another limitation is the need for biological validation of the selected gene sets. Although the
SNR-OGSCC method successfully identifies the most informative genes for classification, further

Revolutionizing cancer classification: the snr-ogscc method for improved gene … (Sara Haddou Bouazza)
470  ISSN: 2252-8938

experimental work is needed to confirm their biological relevance. Validation studies would be critical before
these gene sets can be used as biomarkers in clinical settings. Additionally, the method’s reliance on machine
learning algorithms tailored to high-dimensional data, such as SVM and KNN, could present challenges in
certain applications, particularly where data distributions deviate from model assumptions.

3.5. Discussion
This study aimed to enhance cancer classification accuracy by employing the SNR-OGSCC
methodology. While prior research has explored various machine learning and statistical techniques to classify
cancer types, they often fall short in balancing classification accuracy with the dimensionality of gene
expression data. For instance, Mallick et al. [34] and Nirmalakumari et al. [35] achieved high classification
accuracies of 98.2% and 98.5% for leukemia, respectively, yet they do not specifically address the need for
reduced gene sets to facilitate interpretability and clinical applicability.
Our findings indicate that the SNR-OGSCC method successfully achieved 100% accuracy across
multiple cancer datasets, including leukemia, colon cancer, and lung cancer, with significantly reduced
numbers of genes selected. For instance, in the leukemia dataset, only 4-5 genes were necessary to achieve
optimal performance, contrasting with Mallick et al. [34], who required 13 genes to reach 98.2% accuracy.
This demonstrates the method's capacity to isolate the most relevant features effectively.
When comparing our results with those of Shafi et al. [8], who reported an accuracy of 95.16% for
colon cancer using random forest techniques, our SNR-OGSCC method surpassed this accuracy at 98.4% while
employing a similar number of genes. This trend continued across other cancer types. For instance, while
Fathi et al. [36] achieved 95% accuracy for lung cancer, our method provided perfect classification (100%)
with fewer gene inputs. Similarly, studies like Rajaguru et al. [9] and Alshareef et al. [37] reached accuracies
of 96.46% and 97.19% for prostate cancer, respectively; our methodology achieved 99.1% accuracy using a
minimal gene subset.
In the lymphoma cancer dataset, Olaniran and Abdullah [38] reported an accuracy of 94.92% with a
hybrid variational bayes (VB) approach, while our method achieved 98.3%, indicating a substantial
improvement. Similarly, for CNS tumors, Painuli et al. [39] obtained an impressive accuracy of 99.6% using
an logistic regression (LR)-based model, but our approach maintained high performance (99.7%) while
simplifying the gene set. In ovarian cancer, Prabhakar and Lee [40] achieved a high classification accuracy of
99.48% using SVM-RBF with genetic bee colony optimization (GBCO). However, our methodology not only
matched this performance but also consistently required fewer genes for classification, demonstrating its
efficiency and effectiveness.
Despite the encouraging results, this study's limitations should be noted. The relatively small sample
sizes in datasets, such as CNS tumors, where Painuli et al. [39] reported an accuracy of 99.6%, may affect the
generalizability of our findings. Furthermore, the study's focus on classification accuracy may overlook other
essential factors, such as precision and recall, which are vital for clinical relevance. The requirement for
biological validation of the selected genes remains another challenge, as noted in various studies, including
those by Prabhakar and Lee [40], which emphasize the need for further experimental confirmation of
computational predictions.
Future research should focus on validating the SNR-OGSCC methodology across larger, more diverse
datasets to confirm its efficacy and reliability in different clinical settings. Investigating the biological
significance of the selected genes could also enhance the method’s applicability in personalized medicine.
Additionally, exploring the integration of other machine learning algorithms, as seen in [41], [42], could further
optimize classification accuracy and robustness.
In summary, our findings provide compelling evidence that the SNR-OGSCC methodology
significantly enhances cancer classification accuracy while reducing the dimensionality of gene expression
data. This approach addresses critical gaps in the existing literature by facilitating the identification of
biologically relevant markers with potential clinical significance. The ability to achieve high classification
accuracy with fewer genes positions SNR-OGSCC as a valuable tool in cancer diagnosis and treatment,
warranting further exploration and validation in future studies.

4. CONCLUSION
The SNR-OGSCC methodology demonstrates significant advancements in the field of cancer
diagnostics through enhanced classification accuracy and reduced dimensionality in gene expression data. Our
study revealed that SNR-OGSCC consistently outperformed traditional methods across multiple cancer types,
achieving remarkable accuracies, including 100% for leukemia, 98.4% for colon cancer, 99.1% for prostate
cancer, and 100% for lung cancer. Additionally, it attained 98.3% for lymphoma cancer, 99.7% for CNS
tumors, and 100% for ovarian cancer, while requiring significantly fewer genes for effective classification—

Int J Artif Intell, Vol. 14, No. 1, February 2025: 466-472

Int J Artif Intell ISSN: 2252-8938  471

e.g., as low as 4 genes for leukemia and ovarian cancer. These improvements highlight the strength of
integrating advanced gene selection and clustering techniques, enabling better separation between cancerous
and non-cancerous samples. The results indicate that by optimizing gene selection through clustering, we can
not only improve the robustness of cancer classification but also facilitate the identification of biologically
relevant biomarkers. For instance, in the leukemia dataset, our method required only 4–5 genes to achieve
perfect accuracy, demonstrating its capacity to retain only the most pertinent features. The potential
implications of our findings extend beyond improved diagnostic capabilities. The SNR-OGSCC method paves
the way for personalized treatment strategies by enabling the identification of specific genetic markers
associated with different cancer types. As such, our methodology serves as a vital tool for researchers and
clinicians striving to enhance cancer detection and treatment. However, to fully realize the potential of the
SNR-OGSCC approach, further validation is required. Future studies should focus on applying this
methodology to larger datasets and diverse cancer types to assess its generalizability and robustness.
Additionally, biological validation of the selected gene sets will be essential to confirm their relevance in
clinical settings. Ultimately, this research contributes to the ongoing effort to refine cancer classification
techniques and emphasizes the importance of integrating computational methods with biological insights for
improved patient outcomes.

REFERENCES
[1] R. H. Elden, V. F. Ghonim, M. M. A. Hadhoud, and W. Al-Atabany, “Transcriptomic marker screening for evaluating the mortality
rate of pediatric sepsis based on Henry gas solubility optimization,” Alexandria Engineering Journal, vol. 68, pp. 693–707, Apr.
2023, doi: 10.1016/j.aej.2022.12.027.
[2] M. Elloumi, M. A. Ahmad, A. H. Samak, A. M. Al-Sharafi, D. Kihara, and A. I. Taloba, “Error correction algorithms in non-null
aspheric testing next generation sequencing data,” Alexandria Engineering Journal, vol. 61, no. 12, pp. 9819–9829, Dec. 2022, doi:
10.1016/j.aej.2022.03.041.
[3] A. Akgül, S. H. A. Khoshnaw, and H. M. Rasool, “Minimizing cell signalling pathway elements using lumping parameters,”
Alexandria Engineering Journal, vol. 59, no. 4, pp. 2161–2169, Aug. 2020, doi: 10.1016/j.aej.2020.01.041.
[4] Y. Esmaeili et al., “Exploring the evolution of tissue engineering strategies over the past decade: from cell-based strategies to gene-
activated matrix,” Alexandria Engineering Journal, vol. 81, pp. 137–169, Oct. 2023, doi: 10.1016/j.aej.2023.08.080.
[5] M. Shaheen, N. Naheed, and A. Ahsan, “Relevance-diversity algorithm for feature selection and modified Bayes for prediction,”
Alexandria Engineering Journal, vol. 66, pp. 329–342, Mar. 2023, doi: 10.1016/j.aej.2022.11.002.
[6] L. Zhang, L. Li, M. Tang, Y. Huan, X. Zhang, and X. Zhe, “A new approach to diagnosing prostate cancer through magnetic
resonance imaging,” Alexandria Engineering Journal, vol. 60, no. 1, pp. 897–904, Feb. 2021, doi: 10.1016/j.aej.2020.10.018.
[7] S. J. Susmi, H. K. Nehemiah, A. Kannan, and J. Christopher, “Relevant gene selection and classification of leukemia gene
expression data,” in Emerging Research in Computing, Information, Communication and Applications, Singapore: Springer, 2016,
pp. 503–510, doi: 10.1007/978-981-10-0287-8_47.
[8] A. S. M. Shafi, M. M. I. Molla, J. J. Jui, and M. M. Rahman, “Detection of colon cancer based on microarray dataset using machine learning
as a feature selection and classification techniques,” SN Applied Sciences, vol. 2, no. 7, Jul. 2020, doi: 10.1007/s42452-020-3051-2.
[9] K. Nirmalakumari, H. Rajaguru, and P. Rajkumar, “Microarray prostate cancer classification using eminent genes,” in 2021 Smart
Technologies, Communication and Robotics (STCR), Oct. 2021, pp. 1–5, doi: 10.1109/STCR51658.2021.9588811.
[10] E. Badr, S. Almotairi, M. A. Salam, and H. Ahmed, “New sequential and parallel support vector machine with grey wolf optimizer for
breast cancer diagnosis,” Alexandria Engineering Journal, vol. 61, no. 3, pp. 2520–2534, Mar. 2022, doi: 10.1016/j.aej.2021.07.024.
[11] A. S. Alzahrani, R. A. Shah, Y. Qian, and M. Ali, “A novel method for feature learning and network intrusion classification,”
Alexandria Engineering Journal, vol. 59, no. 3, pp. 1159–1169, Jun. 2020, doi: 10.1016/j.aej.2020.01.021.
[12] T. Althobaiti, S. Althobaiti, and M. M. Selim, “An optimized diabetes mellitus detection model for improved prediction of accuracy
and clinical decision-making,” Alexandria Engineering Journal, vol. 94, pp. 311–324, May 2024, doi: 10.1016/j.aej.2024.03.044.
[13] F. Noman et al., “Multistep short-term wind speed prediction using nonlinear auto-regressive neural network with exogenous
variable selection,” Alexandria Engineering Journal, vol. 60, no. 1, pp. 1221–1229, Feb. 2021, doi: 10.1016/j.aej.2020.10.045.
[14] N. M. Sallam, A. I. Saleh, H. Arafat Ali, and M. M. Abdelsalam, “An efficient EGWO algorithm as feature selection for B-ALL
diagnoses and its subtypes classification using peripheral blood smear images,” Alexandria Engineering Journal, vol. 68, pp. 39–
66, Apr. 2023, doi: 10.1016/j.aej.2023.01.004.
[15] A. M. Khalid, W. Said, M. Elmezain, and K. M. Hosny, “A new binary object-oriented programming optimization algorithm for
solving high-dimensional feature selection problem,” Alexandria Engineering Journal, vol. 85, pp. 72–85, Dec. 2023, doi:
10.1016/j.aej.2023.11.021.
[16] D. M. Mahalakshmi and S. Sumathi, “Brain tumour segmentation strategies utilizing mean shift clustering and content based active
contour segmentation,” ICTACT Journal on Image and Video Processing, vol. 9, no. 4, pp. 2002–2008, May 2019, doi:
10.21917/ijivp.2019.0284.
[17] A. S. Negm, O. A. Hassan, and A. H. Kandil, “A decision support system for acute leukaemia classification based on digital
microscopic images,” Alexandria Engineering Journal, vol. 57, no. 4, pp. 2319–2332, Dec. 2018, doi: 10.1016/j.aej.2017.08.025.
[18] E. E. Nithila and S. S. Kumar, “Segmentation of lung nodule in CT data using active contour model and fuzzy C-mean clustering,”
Alexandria Engineering Journal, vol. 55, no. 3, pp. 2583–2588, 2016, doi: 10.1016/j.aej.2016.06.002.
[19] Y. Abo-Elnaga and S. Nasr, “K-means cluster interactive algorithm-based evolutionary approach for solving bilevel multi-objective
programming problems,” Alexandria Engineering Journal, vol. 61, no. 1, pp. 811–827, Jan. 2022, doi: 10.1016/j.aej.2021.04.098.
[20] H. Wu, Y. Chen, W. Zhu, Z. Cai, A. A. Heidari, and H. Chen, “Feature selection in high-dimensional data: an enhanced RIME
optimization with information entropy pruning and DBSCAN clustering,” International Journal of Machine Learning and
Cybernetics, vol. 15, no. 9, pp. 4211–4254, Sep. 2024, doi: 10.1007/s13042-024-02143-1.
[21] E. Guven, “Decision of the optimal rank of a nonnegative matrix factorization model for gene expression data sets utilizing the unit
invariant knee method: development and evaluation of the elbow method for rank selection,” JMIR Bioinformatics and
Biotechnology, vol. 4, Jun. 2023, doi: 10.2196/43665.

Revolutionizing cancer classification: the snr-ogscc method for improved gene … (Sara Haddou Bouazza)
472  ISSN: 2252-8938

[22] H. E. Saroğlu et al., “Machine learning, IoT and 5G technologies for breast cancer studies: a review,” Alexandria Engineering
Journal, vol. 89, pp. 210–223, Feb. 2024, doi: 10.1016/j.aej.2024.01.043.
[23] M. Roshani et al., “Evaluation of flow pattern recognition and void fraction measurement in two phase flow independent of oil pipeline’s
scale layer thickness,” Alexandria Engineering Journal, vol. 60, no. 1, pp. 1955–1966, Feb. 2021, doi: 10.1016/j.aej.2020.11.043.
[24] I. G. P. S. Wijaya, I. B. K. Widiartha, F. Bimantoro, and A. W. Septiadi, “Buildings cracks classification using zoning and invariant
moment features and quadratic discriminant analysis classifier,” Lontar Komputer: Jurnal Ilmiah Teknologi Informasi, Dec. 2019,
doi: 10.24843/LKJITI.2019.v10.i03.p04.
[25] T. Mahesh, A. Kaladevi, J. Balajee, V. Vivek, M. Prabu, and V. Muthukumaran, “An efficient ensemble method using k-fold cross
validation for the early detection of benign and malignant breast cancer,” International Journal of Integrated Engineering, vol. 14,
no. 7, Dec. 2022, doi: 10.30880/ijie.2022.14.07.015.
[26] H. A. Afify, “Evaluation of change detection techniques for monitoring land-cover changes: a case study in new Burg El-Arab
area,” Alexandria Engineering Journal, vol. 50, no. 2, pp. 187–195, Jun. 2011, doi: 10.1016/j.aej.2011.06.001.
[27] T. R. Golub et al., “Molecular classification of cancer: class discovery and class prediction by gene expression monitoring,” Science,
vol. 286, no. 5439, pp. 531–537, Oct. 1999, doi: 10.1126/science.286.5439.531.
[28] C. Park and S.-B. Cho, “Evolutionary ensemble classifier for lymphoma and colon cancer classification,” in The 2003 Congress on
Evolutionary Computation, 2003. CEC ’03., vol. 4, pp. 2378–2385, doi: 10.1109/CEC.2003.1299385.
[29] A. H. Chen, Y.-W. Tsau, and C.-H. Lin, “Novel methods to identify biologically relevant genes for leukemia and prostate cancer
from gene expression profiles,” BMC Genomics, vol. 11, no. 1, 2010, doi: 10.1186/1471-2164-11-274.
[30] G. J. Gordon et al., “Translation of microarray data into clinically relevant cancer diagnostic tests using gene expression ratios in
lung cancer and mesothelioma,” Cancer Research, vol. 62, no. 17, pp. 4963–4967, 2002.
[31] M. A. Shipp et al., “Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine
learning,” Nature Medicine, vol. 8, no. 1, pp. 68–74, Jan. 2002, doi: 10.1038/nm0102-68.
[32] J.-Y. Yeh, “Applying data mining techniques for cancerclassification on gene expression data,” Cybernetics and Systems, vol. 39,
no. 6, pp. 583–602, Aug. 2008, doi: 10.1080/01969720802188292.
[33] E. F. Petricoin et al., “Use of proteomic patterns in serum to identify ovarian cancer,” The Lancet, vol. 359, no. 9306, pp. 572–577,
Feb. 2002, doi: 10.1016/S0140-6736(02)07746-2.
[34] P. K. Mallick, S. K. Mohapatra, G.-S. Chae, and M. N. Mohanty, “Convergent learning–based model for leukemia classification from
gene expression,” Personal and Ubiquitous Computing, vol. 27, no. 3, pp. 1103–1110, Jun. 2023, doi: 10.1007/s00779-020-01467-3.
[35] K. Nirmalakumari, H. Rajaguru, and P. Rajkumar, “Leukemia cancer classification using extrusive genes from microarray data,”
Second International Conference on Circuits, Signals, Systems and Securities (ICCSSS - 2022), vol. 2725, no. 1, 2023, doi:
10.1063/5.0125232.
[36] H. Fathi, H. AlSalman, A. Gumaei, I. I. M. Manhrawy, A. G. Hussien, and P. El-Kafrawy, “An efficient cancer classification model
using microarray and high‐dimensional data,” Computational Intelligence and Neuroscience, vol. 2021, no. 1, Jan. 2021, doi:
10.1155/2021/7231126.
[37] A. M. Alshareef et al., “Optimal deep learning enabled prostate cancer detection using microarray gene expression,” Journal of
Healthcare Engineering, vol. 2022, pp. 1–12, Mar. 2022, doi: 10.1155/2022/7364704.
[38] O. R. Olaniran and M. A. A. Abdullah, “Subset selection in high-dimensional genomic data using hybrid variational bayes and
bootstrap priors,” Journal of Physics: Conference Series, vol. 1489, no. 1, Mar. 2020, doi: 10.1088/1742-6596/1489/1/012030.
[39] D. Painuli, S. Bhardwaj, and U. Kose, “Optimized diagnosis of central nervous system (CNS) cancer using gene expression
microarray & machine learning (ML) methods,” European Chemical Bulletin, vol. 12, no. 10, pp. 9757-9771, 2023.
[40] S. K. Prabhakar and S.-W. Lee, “An integrated approach for ovarian cancer classification with the application of stochastic
optimization,” IEEE Access, vol. 8, pp. 127866–127882, 2020, doi: 10.1109/ACCESS.2020.3006154.
[41] A. Al-Murad and M. F. Hossain, “An integrated feature selection method for neural network to classify ovarian cancer,” in 2021
International Conference on Automation, Control and Mechatronics for Industry 4.0 (ACMI), Jul. 2021, pp. 1–6, doi:
10.1109/ACMI53878.2021.9528106.
[42] M. Z. H. Akmal, and D. Fitrianah, “Exploring feature selection for microarray classification,” International Journal of Computing
and Digital Systems, vol. 16, no. 1, 2024.

BIOGRAPHIES OF AUTHORS

Sara Haddou Bouazza holds a Doctorate in Electrical Engineering and Informatics,

as well as a Master's in Electrical Engineering from Cadi Ayyad University, Marrakech. She also
completed her Bachelor's in Physical Sciences. Currently, she is a professor and researcher at the
LAMIGEP laboratory, EMSI Marrakech. Her research includes AI techniques for cancer
classification, gene expression analysis, and security challenges in IoT environments. She has
published numerous papers, including recent work on leukemia classification and AI in CNS
tumors. She can be contacted at email: [email protected].

Jihad Haddou Bouazza is an engineer specializing in software engineering and image

processing from IGA Institut Supérieur du Génie Appliqué, Marrakech. Currently, he serves as a
senior full stack developer & tech lead at Nexular Corp. He is certified in Python, machine learning,
and as a certified network security specialist (CNSS). His research includes pattern recognition
using artificial intelligence, with a publication presented at the GAST24 congress. He can be
contacted at email: [email protected].

Int J Artif Intell, Vol. 14, No. 1, February 2025: 466-472