A comparative analysis of swarm intelligence techniques for feature selection in cancer classification

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning in Healthcare

Background:

• Feature selection is crucial for identifying informative genes from high-dimensional microarray data in cancer classification.
• Existing swarm intelligence (SI) techniques like Particle Swarm Optimization (PSO), Cuckoo Search (CS), and Shuffled Frog Leaping (SFL) are used but can face premature convergence.
• The signal-to-noise ratio (SNR), T-statistics, and F-test are common methods for initial gene ranking.

Purpose of the Study:

• To propose and evaluate a novel swarm intelligence technique, shuffled frog leaping with Lévy flight (SFLLF), for enhanced gene feature selection in cancer classification.
• To improve the convergence properties of the SFL algorithm by incorporating Lévy flight.
• To compare the performance of SFLLF against other SI techniques (PSO, CS, SFL) for identifying informative genes.

Main Methods:

• Gene expression data from 10 benchmark cancer datasets were utilized.
• Feature selection was performed using PSO, CS, SFL, and the proposed SFLLF algorithm.
• The selected informative genes were used to train a k-nearest neighbor (k-NN) classifier.
• Performance was evaluated based on classification accuracy.

Main Results:

• The SFLLF feature selection method, when combined with the k-NN classifier, demonstrated superior performance compared to PSO, CS, and SFL.
• The incorporation of Lévy flight in SFLLF helped to avoid premature convergence, leading to more robust feature selection.
• Consistent improvements in classification accuracy were observed across the 10 benchmark datasets.

Conclusions:

• The proposed SFLLF algorithm is an effective and efficient method for gene feature selection in cancer classification.
• SFLLF offers an advantage over traditional SI methods by mitigating premature convergence issues.
• This approach holds promise for improving diagnostic accuracy and understanding of cancer through bioinformatics analysis.