A hierarchical learning based artificial bee colony algorithm for numerical global optimization and its applications

The Artificial Bee Colony algorithm (ABC) is a swarm intelligence algorithm inspired by honey bee harvesting behavior. It boasts the benefits of minimal parameters and strong exploration capabilities. However, the ABC algorithm is still susceptible to local optima entrapment and lacks consideration of selection probability in the onlooker bee phase, leading to reduced convergence accuracy in later search stages. To address these issues, this paper introduces an enhanced ABC algorithm called Hierarchical Learning-based Artificial Bee Colony (HLABC). Initially, a hierarchical learning approach is devised, dividing the entire population into distinct layers based on solution quality. In this hierarchical approach, bees at lower layers can access much better advantageous information from higher layers. Secondly, the exploitation ability of onlooker bees is enhanced through novel strategies designed based on hierarchical learning. Thirdly, the exploration ability of scout bees is strengthened by implementing an opposition-based learning method. To evaluate the performance of the proposed algorithm, 69 benchmark functions from four benchmark suites (CEC2005, CEC2010, CEC2013 and CEC2022) are used to test the performance of HLABC, along with five variants of the ABC algorithm, The experimental statistical results show that the HLABC algorithm outperforms the ABC algorithm on all test problems with an average winning rate of 89%. Furthermore, to validate the performance of the HLABC algorithm in real-world optimization problems, this paper applies the HLABC algorithm to two practical applications: the deployment of wireless sensor networks (WSNs), the power scheduling problem in a smart home (PSPSH) and the multi-thresholding image segmentation (MIS). The experimental and statistical results demonstrate that HLABC is an efficient and stable optimizer. It shows better or comparable performance compared to other ABC variants when considering the quality of solutions for a suite of benchmark problems and real-world optimization problems. These findings affirm the effectiveness and versatility of the HLABC algorithm in addressing both theoretical and practical optimization challenges.