An Efficient Improved Greedy Harris Hawks Optimizer and Its Application to Feature Selection.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Harris Hawks Optimization (HHO) exhibits limitations in exploration-exploitation switching flexibility and exploitation efficiency.
• Feature Selection (FS) is crucial for reducing data dimensionality and improving model performance.

Purpose of the Study:

• To propose an Improved Greedy Harris Hawks Optimizer (IGHHO) addressing HHO's limitations.
• To apply IGHHO to the Feature Selection (FS) problem, including data imbalance challenges.

Main Methods:

• IGHHO incorporates a novel transformation strategy for flexible search-exploitation switching and improved global search capability.
• The original HHO exploitation phase is replaced with improved differential perturbation and a greedy strategy.
• New objective functions are developed for FS problems with data imbalance.

Main Results:

• IGHHO demonstrated superior performance against seven algorithms on CEC2017 benchmark functions.
• IGHHO outperformed comparison algorithms in classification accuracy and reduced feature subset length for imbalanced data FS.
• The algorithm effectively handles global optimization across diverse feature functions and practical applications.

Conclusions:

• IGHHO offers enhanced flexibility and efficiency for optimization problems.
• The proposed method is effective for feature selection, particularly in scenarios with imbalanced datasets.
• IGHHO shows significant potential for both theoretical optimization and real-world applications.