A Revised Take on the Bee Optimization Algorithm Through Diverse Initialization, Adaptive Neighbors, Global Tracking, Gradual Reduction with Balanced Exploitation for Better Results
A Revised Take on the Bee Optimization Algorithm Through Diverse Initialization, Adaptive Neighbors, Global Tracking, Gradual Reduction with Balanced Exploitation for Better Results

Author(s): Hiteshkumar Nimbark, Bhumit Jograna, Sparsh Nimbark
Subject(s): Methodology and research technology, ICT Information and Communications Technologies
Published by: Scientia Moralitas Research Institute
Keywords: Artificial Bee Colony (ABC) Algorithm; Swarm Intelligence; Metaheuristic Optimization; Enhanced Bee Algorithm; Search Space Exploration and Exploitation; Dynamic Parameter Adaptation; Adaptive Neighborhood Search
Summary/Abstract: The Bee Algorithm is a well-known swarm intelligence technique inspired by the foraging behavior of honeybees. Despite its success in solving various optimization problems, the standard version of the algorithm is often limited by several inherent weaknesses. These include poor diversity during the initial population setup, a fixed neighborhood search strategy that lacks adaptability, and the tendency to converge prematurely to suboptimal solutions. Additionally, many existing implementations fail to retain the best-found solution across iterations, leading to a drop in final solution quality. This paper introduces a modified version of the Bee Algorithm that addresses these issues through five targeted enhancements. The first involves a diversified initialization method that systematically distributes the initial population across the search space to prevent clustering and encourage broader exploration. The second introduces an adaptive neighborhood search radius that evolves with the number of iterations, providing a smooth transition from global search to local refinement. Third, a global best tracking mechanism is implemented to ensure the most optimal solution is retained throughout the process. Fourth, a gradual reduction strategy for the search radius prevents overly rapid convergence and maintains diversity for a longer period. Finally, the update scheme is adjusted to better balance exploitation of elite solutions and the integration of new candidates, which improves both convergence reliability and robustness. Comparative experiments using a set of well-established benchmark functions demonstrate that the proposed improvements consistently outperform the standard Bee Algorithm and several recent variants in terms of convergence speed, accuracy, and stability, without introducing significant computational overhead. The proposed modifications are easy to implement and offer a practical upgrade for applications where reliable global optimization is required.

• Details
• Contents