A feature selection method based on the Golden Jackal-Grey Wolf Hybrid Optimization Algorithm.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• High-dimensional datasets present challenges in feature selection due to redundant, irrelevant, and noisy features.
• Existing optimization algorithms may lack efficiency and accuracy in complex data scenarios.
• Effective dimensionality reduction is crucial for improving model performance and interpretability.

Purpose of the Study:

• To propose a novel hybrid optimization algorithm for effective feature selection.
• To develop a data dimensionality reduction technique for high-dimensional datasets.
• To enhance classification accuracy and algorithm stability through optimized feature selection.

Main Methods:

• A hybrid optimization algorithm, the multi-strategy fusion GJO-GWO, combining Golden Jackal Optimization (GJO) and Grey Wolf Optimizer (GWO) with Lagrange interpolation.
• Evaluation of the GJO-GWO algorithm on eight complex benchmark functions (Case 1).
• Application of the GJO-GWO algorithm to ten feature selection problems (Case 2).

Main Results:

• The GJO-GWO algorithm demonstrated superior optimization performance compared to other methods.
• Consistent improvements were observed in smaller means and lower standard deviations across benchmark functions and feature selection tasks.
• The algorithm achieved higher classification accuracy and reduced execution times.

Conclusions:

• The proposed GJO-GWO algorithm offers superior optimization capabilities for feature selection.
• The hybrid approach provides enhanced classification accuracy and stability for high-dimensional data.
• This method represents an effective strategy for data dimensionality reduction and feature optimization.