Artificial bee colony algorithm based on multi-neighbor guidance
Xinyu Zhou,Guisen Tan,Hui Wang,Yong Ma,Shuixiu Wu
DOI: https://doi.org/10.1016/j.eswa.2024.125283
IF: 8.5
2024-09-11
Expert Systems with Applications
Abstract:In recent years, to enhance the exploitation capability of the artificial bee colony (ABC) algorithm, some neighborhood-based ABC variants have been designed by utilizing the best neighbor. However, they may suffer from an issue that the algorithm easily becomes too greedy and further causes premature convergence. In fact, from the perspective of individual information, an individual contains not only fitness information but also distance information, and both kinds of information can be utilized. Accordingly, in the scenario of neighborhood topology, not only the neighbors in terms of fitness, but also the neighbors considering distance can be used to guide the search. Motivated by these, this paper proposes a new neighborhood-based ABC variant with multi-neighbor guidance (ABC-MNG). In ABC-MNG, the neighbor with the best fitness value and the neighbors with the farthest and nearest Euclidean distances are chosen. Subsequently, based on the three neighbors, three different offspring are generated, and a similarity metric is designed to evaluate these offspring, aiming to avoid a heavy computational burden. Meanwhile, to preserve the search experience of the scout bee, a modified neighborhood search operator is presented using the multiple neighbors. Extensive experiments are carried out on two well-known test suites and one real-world problem. Ten advanced ABC variants, including five neighborhood-based ABC variants and five other well-established ABC variants, are compared with ABC-MNG, and the results confirm that ABC-MNG is highly competitive, surpassing its competitors.
computer science, artificial intelligence,engineering, electrical & electronic,operations research & management science