A novel firefly algorithm approach for efficient feature selection with COVID-19 dataset.

Area of Science:

• Computer Science
• Machine Learning
• Data Science

Background:

• High-dimensional datasets often contain irrelevant or redundant features, hindering machine learning model performance.
• Effective feature selection is crucial for improving classification accuracy and deriving meaningful conclusions from data.

Purpose of the Study:

• To propose a novel, enhanced version of the firefly algorithm (FA) specifically adapted for the feature selection challenge.
• To evaluate the performance of the proposed FA against existing metaheuristics on benchmark and real-world datasets.

Main Methods:

• Development of a novel variant of the firefly algorithm (FA) for feature selection.
• Validation on standard unconstrained benchmarks.
• Application and testing on 21 University of California, Irvine (UCI) datasets, a COVID-19 patient health prediction dataset, and a microcontroller microarray dataset.

Main Results:

• The proposed FA variant demonstrated significant performance improvements over the basic FA and other state-of-the-art metaheuristics.
• Achieved the best classification accuracy on 13 out of 21 UCI datasets.
• Showcased robustness and efficiency in convergence, solution quality, and classification accuracy across all tested practical applications.

Conclusions:

• The novel firefly algorithm is a robust and efficient method for feature selection in machine learning.
• The proposed approach significantly enhances classification accuracy, outperforming existing methods on complex, real-world datasets.