A multi-strategy framework for enhancing Harris hawks optimization for global optimization problems
Abstract
The standard Harris Hawks Optimization (HHO) algorithm often suffers from premature convergence and stagnation on complex, high-dimensional problems, particularly in wrapper-based feature selection (FS) for medical diagnosis, where costly fitness evaluations demand efficient search strategies. This paper proposes a Modified HHO (MHHO) that integrates three synergistic mechanisms: (i) Leader-Guided Perching (LGP), which introduces leader bias during exploration to accelerate movement toward promising regions; (ii) an Adaptive Deception Factor (ADF), which dynamically scales Lévy-flight intensity to escape deceptive local optima; and (iii) a Hierarchical Attack Strategy (HAS), which replaces the hard besiege phase with a mentor-guided exploitation step to mitigate swarm clumping and refine local search. A dual-domain evaluation was conducted. In Experiment Series 1, MHHO statistically outperformed the standard HHO on 18 of 23 benchmark functions, at the expected cost of higher computational time. In Experiment Series 2, involving wrapper-based FS on 15 medical datasets, MHHO achieved higher mean classification accuracy on most datasets while exhibiting enhanced stability, resulting in a favorable accuracy–sparsity trade-off through minimal yet high-performing feature subsets. Friedman’s test confirmed a significant global difference (p = 0.0013), and Holm’s post-hoc analysis verified that LGP, ADF, and HAS each contribute significant improvements over standard HHO. Overall,…
Read the paper
Related studies
- Quantum dots: synthesis, bioapplications, and toxicity
- A Guide to In Silico Drug Design
- In silico toxicology protocols
- Automated in Silico Design of Homogeneous Catalysts
- In-silico strategies in nano-drug design: Bridging nanomaterials and pharmacological applications
- mRNA codon optimization with quantum computers
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.