Prey capture enhanced Harris hawks optimizer for wrapper-based feature selection in high-dimensional medical data.

Area of Science:

• Optimization algorithms
• Computational intelligence
• Data science

Background:

• The Harris Hawks Optimizer (HHO) is effective for feature selection but struggles with high-dimensional data, local optima, and computational cost.
• Limitations in HHO's prey capture mechanism can hinder its performance on complex datasets.

Purpose of the Study:

• Introduce an enhanced HHO algorithm, the Prey Capture Harris Hawks Optimizer (PCHHO), to improve prey-capture ability.
• Develop a binary variant (bPCHHO) for wrapper-based feature selection on high-dimensional datasets.

Main Methods:

• PCHHO integrates crossover and mutation operators to boost exploratory-exploitative capabilities.
• Evaluated PCHHO on the CEC2017 benchmark suite against HHO and other metaheuristics.
• bPCHHO was tested on 15 high-dimensional medical datasets against six binary metaheuristics.

Main Results:

• PCHHO demonstrated superior performance on the CEC2017 benchmark suite.
• bPCHHO achieved a 77% reduction in classification error, 8% reduction in computational time, and 73% fewer features selected compared to bHHO.
• Statistical tests (Wilcoxon signed-rank, Friedman) confirmed the significant performance improvements.

Conclusions:

• PCHHO and bPCHHO show excellent performance in benchmark optimization and feature selection.
• bPCHHO is highly effective for wrapper-based feature selection on high-dimensional medical data.
• The enhanced algorithms offer promising potential for practical applications in data analysis.