Enhanced Feature Selection Based on Integration Containment Neighborhoods Rough Set Approximations and Binary Honey...

Area of Science:

• Data Science
• Machine Learning
• Artificial Intelligence

Background:

• Traditional rough set approximations (RSA) have limitations in defining precise boundaries.
• Feature selection (FS) is crucial for improving classification model performance.
• Optimization algorithms are essential for efficient feature selection.

Purpose of the Study:

• Introduce a novel rough set approximation model, CRSA, to generalize RSA and reduce boundary regions.
• Integrate CRSA with the binary honey badger optimization (HBO) algorithm for feature selection.
• Evaluate the effectiveness of the CRSA-based HBO approach in enhancing classification accuracy.

Main Methods:

• Developed a novel rough set approximation model based on containment neighborhoods (CRSA).
• Implemented a binary version of the honey badger optimization (HBO) algorithm for feature selection.
• Evaluated the CRSA-based HBO approach on ten benchmark datasets, comparing it with existing FS methods.

Main Results:

• CRSA demonstrated superiority over traditional RSA by effectively minimizing boundary areas.
• The CRSA-based HBO approach significantly improved classification accuracy across all tested datasets.
• BHBO outperformed other well-known feature selection methods in terms of performance metrics.

Conclusions:

• CRSA offers a valuable generalization of traditional RSA with improved boundary reduction.
• The integration of CRSA with HBO provides an effective feature selection strategy.
• This novel approach enhances classification performance and demonstrates high potential in machine learning applications.