Automatic programming is a branch of artificial intelligence that presents each solution as a mathematical formula based on heuristic mechanisms. In this study, artificial bee colony expression programming (ABCEP) is presented, which is combined simultaneously with expression programming. By using expression sharing to generate new solutions, the proposed method can minimize certain deficiencies of artificial bee colony programming, such as weak convergence and high locality. A total number of 15 real-world regression benchmark functions was used to evaluate the performance of the proposed model. For comparison purposes, successful run percentage, mean best cost, convergence performance, and run time of ABCEP were compared to those of other tested automatic programming algorithms, including artificial bee colony programming, gene expression programming, genetic programming, and quick artificial bee colony programming. A Wilcoxon signed-rank test was also done to compare the behavior of the algorithms. Additionally, the accuracy of all algorithms was then evaluated using three real-world practical benchmarks. The results indicate that the predictions generated by ABCEP are better than those obtained by other control algorithms based on successful runs, mean fitness values, and convergence rate.

自动编程是人工智能的一个分支，它根据启发式机制将每个解决方案表示为数学公式。在这项研究中，提出了人工蜂群表达编程（ABCEP），它与表达编程同时进行。通过使用表达式共享生成新的解决方案，该方法可以最大程度地减少人工蜂群编程的某些缺陷，例如收敛性较弱和局部性较高。总共使用了15个真实世界的回归基准函数来评估所提出模型的性能。为了进行比较，将ABCEP的成功运行百分比，平均最佳成本，收敛性能和运行时间与其他经过测试的自动编程算法（包括人工蜂群编程）进行了比较，基因表达编程，遗传编程和快速人工蜂群编程。还进行了Wilcoxon符号秩检验来比较算法的行为。此外，然后使用三个实际的实用基准评估了所有算法的准确性。结果表明，基于成功运行，平均适​​应度值和收敛速度，ABCEP生成的预测优于其他控制算法获得的预测。