SGA-Driven feature selection and random forest classification for enhanced breast cancer diagnosis: A comparative study

SGA-Driven feature selection and random forest classification for enhanced breast cancer diagnosis: A comparative study

Abstract In this study, we propose a novel approach for breast cancer classification that integrates the Seagull Optimization Algorithm (SGA) for feature selection with the Random Forest (RF) classifier for effective data classification. The novelty of our approach lies in the first-time application...

Full description

Saved in:
Bibliographic Details
Main Authors: Abrar Yaqoob, Navneet Kumar Verma, Mushtaq Ahmad Mir, Ghanshyam G. Tejani, Nashwa Hassan Babiker Eisa, Hind Mamoun Hussien Osman, Mohd Asif Shah
Format: Article
Language:English
Published: Nature Portfolio 2025-03-01
Series:Scientific Reports
Subjects:
Seagull optimization algorithm
Random forest
Cancer classification
High dimensional data
Online Access:https://doi.org/10.1038/s41598-025-95786-1
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract In this study, we propose a novel approach for breast cancer classification that integrates the Seagull Optimization Algorithm (SGA) for feature selection with the Random Forest (RF) classifier for effective data classification. The novelty of our approach lies in the first-time application of SGA for gene selection in breast cancer diagnosis, where SGA systematically explores the feature space to identify the most informative gene subsets, thereby improving classification accuracy and reducing computational complexity. The selected features are subsequently classified using RF, known for its robustness and high accuracy in handling complex datasets. To evaluate the effectiveness of the proposed method, we compared it with other classifiers, including Linear Regression (LR), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN). The proposed SGA-RF combination achieved a best mean accuracy of 99.01% with 22 genes, outperforming other methods and demonstrating consistent performance across varying feature subsets. The mean accuracies ranged from 85.35 to 94.33%, highlighting a balance between feature reduction and classification accuracy. Future work will explore the integration of other nature-inspired algorithms and deep learning models to further enhance performance and clinical applicability.
ISSN:2045-2322

Similar Items