SFLA IWSSr

SFLA IWSSr performs hybrid filter-wrapper feature selection using Relief, the Shuffled Frog Leaping Algorithm (SFLA), and IWSSr to identify compact gene subsets from high-dimensional gene expression datasets that improve classification accuracy and reduce overfitting.

Key Features:

• Hybrid Filter-Wrapper: Integrates filter and wrapper methodologies to combine the complementary advantages of both approaches for feature selection.
• Metaheuristic Algorithms: Employs the Shuffled Frog Leaping Algorithm (SFLA) and the IWSSr method to perform global and local search for relevant features.
• Two-Phase Implementation: Implements a filter phase using the Relief method to weight and reduce features followed by a wrapper phase using SFLA and IWSSr to refine the reduced feature set.
• Dimensionality Reduction and Overfitting Mitigation: Produces compact gene subsets to reduce dimensionality, computational cost, and overfitting in low-sample-size datasets.
• Classification Performance Improvement: Selects feature subsets intended to enhance downstream classification accuracy.

Scientific Applications:

• Gene expression analysis: Selects minimal informative subsets of genes from high-dimensional gene expression datasets.
• Disease classification: Improves input feature sets for classifiers used in disease diagnosis and subtype discrimination.
• Genomic studies and biomarker discovery: Supports identification of compact gene signatures to study genetic influences on biological processes and potential biomarkers.

Methodology:

The method applies a two-phase pipeline consisting of an initial filtering step using the Relief algorithm to weight features, followed by wrapper optimization using the Shuffled Frog Leaping Algorithm (SFLA) and IWSSr to search and refine optimal feature subsets.