Improved artificial bee colony algorithm based on two-dimensional queue structure for complex optimization problems

Abstract

In the face of increasingly complex optimization problems, the Artificial Bee Colony algorithm still encounters some unresolved issues. For instance, the randomness involved in generating the initial population provides diversity to optimization solutions but also leads to convergence among individuals due to the current pseudo-randomness in this process. To address this challenge, this study introduces an improved ABC algorithm based on a two-dimensional queue structure, referred to as KLABC. The KLABC algorithm introduces several key advantages over traditional ABC algorithms. By employing the K-Means algorithm, it enhances the assessment of initial population individuals, ensuring a higher-quality population that aligns with specified requirements. The integration of a two-dimensional queue-based search strategy, along with the application of the first-in, first-out (FIFO) mechanism, not only preserves population diversity but also minimizes the risk of discarding potential optimal solutions during iterative searches. Augmenting this framework, the concepts of the global best solution (GBest) and queue best solution (QBest) are incorporated to fortify the exploration formula, expediting the algorithm's convergence rate. Experimental results on clustering optimization problems across seven UCI datasets validate the effectiveness and robustness of the proposed algorithm. Experimental results indicate that the KLABC algorithm achieves faster convergence compared to the traditional ABC algorithm, with an average performance improvement of 55.63%. Additionally, when comparing the runtime of different algorithms in handling large-scale data, the KLABC algorithm effectively maintains its performance even in scenarios with larger data sizes.