Improved Black-Winged Kite Algorithm with Multi-Strategy Optimization for Identifying Dendrobium huoshanense.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Original black-winged kite optimization algorithm (BKA) suffers from limited search capability due to low-quality initial populations.
• Reduced population diversity in BKA arises from blind following behavior during migration, hindering optimal exploration-exploitation balance.

Purpose of the Study:

• To enhance the black-winged kite optimization algorithm (BKA) by addressing limitations in initial population quality and population diversity.
• To improve the algorithm's convergence precision, exploration-exploitation balance, and overall search capabilities.
• To validate the enhanced algorithm's effectiveness on benchmark functions and a practical application.

Main Methods:

• Incorporated opposition-based learning for higher-quality initial population generation.
• Integrated a differential mutation strategy during migration to mitigate blind leader-following and enhance information exchange.
• Replaced the absorption boundary method with a random boundary approach to increase population diversity.

Main Results:

• The improved algorithm (BKAIM) demonstrated superior convergence performance and robustness across CEC2017, CEC2019, CEC2021, and CEC2022 benchmark functions.
• Wilcoxon rank-sum test comparisons confirmed BKAIM's enhanced performance against other algorithms.
• BKAIM successfully optimized a support vector machine (SVM) model for grading Dendrobium huoshanense using near-infrared spectral data.

Conclusions:

• The proposed BKAIM significantly overcomes the limitations of the original BKA, offering improved optimization capabilities.
• BKAIM provides a robust and effective solution for complex optimization problems.
• The algorithm's practical applicability is confirmed through successful optimization of an SVM model for plant identification.