Hybrid rice optimization algorithm inspired grey wolf optimizer for high-dimensional feature selection.

Area of Science:

• Computational intelligence
• Bioinformatics
• Machine learning

Background:

• Feature selection (FS) is crucial for dimensionality reduction, but existing metaheuristic algorithms like Grey Wolf Optimizer (GWO) struggle with high-dimensional data due to poor adaptability and diversity.
• The Hybrid Rice Optimization (HRO) algorithm shows promise for finding optimal solutions.

Purpose of the Study:

• To propose a novel hybrid metaheuristic algorithm, HRO-GWO, for effective feature selection.
• To enhance GWO's performance by integrating HRO's optimization capabilities and introducing multi-strategy enhancements.

Main Methods:

• Developed HRO-GWO by incorporating a dynamical regulation strategy for parameter optimization and a multi-strategy co-evolution model (neighborhood search, dual-crossover, selfing) to boost population diversity.
• Implemented a hybrid filter-wrapper framework using chi-square and HRO-GWO for efficient selection of informative features.
• Evaluated performance on benchmark functions and small-sample, high-dimensional biomedical datasets.

Main Results:

• The HRO-GWO algorithm demonstrated improved adaptability, diversity, and accuracy compared to standard GWO on high-dimensional data.
• The hybrid filter-wrapper framework effectively selected pertinent features, enhancing classification performance and reducing computation time.
• Experimental results showed HRO-GWO outperformed state-of-the-art methods on benchmark functions and biomedical datasets.

Conclusions:

• The proposed HRO-GWO algorithm offers a robust and efficient solution for feature selection in high-dimensional data.
• This novel approach significantly enhances classification performance and computational efficiency, particularly in biomedical applications.