Abstract:  This paper presents an improved artificial bee colony algorithm to overcome the weaker search ability and imbalance in computing resource allocation. To enhance the neighborhood search of the elites and global best solution, two new strategies are first designed for the employed bees and onlooker bees, respectively. Then a local learning strategy is used to the onlooker bees chosen probably to speed up the convergence speed and enhance the global optimization ability. Finally, to balance the exploration and exploitation effectively, the crisscross search method is employed to enhance the performance of the onlooker bees and global best solution such that the diversity of the population is preserved and the premature convergence phenomenon is reduced. The proposed algorithm is compared with six excellent meta-heuristic algorithms on 10 classical benchmark functions and 30 CEC2014 benchmark functions with different dimensions. The experimental results show that the proposed algorithm is very competitive.

Key words: artificial bee colony (ABC), crisscross search, local learning, numerical optimization, neighborhood search

摘要： 为克服人工蜂群算法搜索策略的局部搜索能力较弱且计算资源分布不均匀等缺点，提出了一种改进人工蜂群算法。首先对雇佣蜂和瞭望蜂，分别设计了新搜索策略，提高了在精英解和全局最好解邻域内的搜索能力；其次对依概率选取的瞭望蜂，采用局部学习策略，加快了收敛速度并增强了全局寻优能力；最后为平衡全局搜索和局部开发，利用十字交叉搜索增强瞭望蜂和全局最好解的局部搜索能力，维持了种群多样性，从而避免了早熟收敛现象。对10个标准测试函数和30个CEC2014测试函数集进行仿真实验，并与四种人工蜂群算法和两种非人工蜂群算法进行比较，结果表明改进的人工蜂群算法全局寻优能力强且提高了收敛速度和精度。

关键词: 人工蜂群算法（ABC）, 十字交叉搜索, 局部学习, 数值优化, 邻域搜索