Mutable Composite Firefly Algorithm for Microarray-Based Cancer Classification
Microarray-based cancer biomarker detection is one of the popular trends for cancer classification. Though existing approaches have given competing performance in terms of classification accuracy and reduced feature subsets, the classification of different cancer microarray datasets still requires...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
UUM Press
2025-01-01
|
| Series: | Journal of ICT |
| Subjects: | |
| Online Access: | https://e-journal.uum.edu.my/index.php/jict/article/view/25595 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Microarray-based cancer biomarker detection is one of the popular trends for cancer classification. Though existing approaches have given competing performance in terms of classification accuracy and reduced feature subsets, the classification of different cancer microarray datasets still requires improvements. Recently, the swarm-based hybrid algorithms have given significant performance in cancer classification. However, the efficiency of a swarm algorithm is dominated by certain factors such as fitness value, convergence, exploration, and exploitation capabilities. Thus, a swarm-based hybrid approach is proposed for cancer classification with a new variant of the Firefly Algorithm (FA) and Correlation-based Feature Selection (CFS) filter. The slow convergence issue in the FA is resolved by non-fixed size solutions termed as mutable size solutions and a composite position update function is designed for the mutable solutions. In addition, the local optima issue is overcome by the population reinitialisation method. The proposed algorithm, named the CFS-Mutable Composite Firefly Algorithm (CFS-MCFA), is evaluated based on two metrics, namely classification accuracy and genes subset size, using a Support Vector Machine (SVM) classifier. Results show that CFS-MCFA-SVM achieved 100% accuracy with only a few biomarkers for all four cancer microarray datasets, indicating the efficiency and the competing performance of the proposed algorithm in biomarker detection for microarray-based cancer classification. Apart from that, the proposed algorithm would also contribute to cancer-related issues upon verifying the relevancy of particular genes via technical analysis from a medical perspective and would be utilised in feature selection applications.
|
|---|---|
| ISSN: | 1675-414X 2180-3862 |