Feature Selection via Chaotic Antlion Optimization.

Area of Science:

• Machine Learning
• Bioinformatics
• Optimization Algorithms

Background:

• Feature selection is crucial for high-dimensional data, especially in biology where numerous biomarkers are generated.
• Optimizing feature selection involves balancing model performance with the number of features used, often as a biobjective problem.

Purpose of the Study:

• To propose a novel optimization approach for feature selection using a chaotic antlion optimizer.
• To enhance the exploration-exploitation tradeoff in multi-objective optimization for improved feature selection.

Main Methods:

• A chaotic version of the antlion optimizer (AO) algorithm was developed, incorporating chaotic maps to refine the exploration-exploitation balance.
• The parameter controlling the random walk range was iteratively adjusted to manage the exploration rate during optimization.
• The methodology was validated on diverse datasets, including ten biological datasets and other data types.

Main Results:

• The chaotic AO demonstrated an improved tradeoff between exploration and exploitation compared to standard AO.
• Evaluation across various datasets showed competitive or superior performance against particle swarm optimization and genetic algorithm variants.
• The approach effectively addressed challenges like premature convergence and local minima trapping.

Conclusions:

• The proposed chaotic antlion optimizer offers a robust method for effective feature selection in high-dimensional datasets.
• This technique holds significant potential for improving classification accuracy and reducing feature dimensionality in biological and other data analyses.