A novel artificial bee colony algorithm integrating K-means clustering and Q-learning

The Artificial Bee Colony (ABC) algorithm is a stochastic, population-based optimization method. While the randomness in generating the initial population ensures solution diversity, it can also result in poor neighborhood structures due to the lack of spatial guidance. To address this challenge, this study propose a novel ABC algorithm integrating K-means clustering and Q-learning, called KQABC. First, the initial population is randomly generated, and then K-means clustering is applied to partition it into multiple clusters, where individuals in the same cluster are considered neighbors. Second, the neighborhood concept is incorporated into both the employed bee and onlooker bee phases. In the employed bee phase, each bee learns from the best individual in its neighborhood, enabling targeted local exploration and avoiding random perturbations. In the onlooker bee phase, Q-learning adaptively selects search strategies. To evaluate the performance of KQABC, we compare it with the original ABC and several state-of-the-art ABC variants on 22 benchmark functions and CEC 2014 test suit. Experimental results demonstrate that KQABC outperforms, or is at least competitive with, existing ABC variants in terms of solution quality, robustness, and convergence speed.