![](https://csdnimg.cn/release/download_crawler_static/15615714/bg4.jpg)
172 L. Cui et al. / Information Sciences 417 (2017) 169–185
Fig. 1. Framework of the original ABC.
search equation in the onlooker bee phase, which exploits the information of the best solution among the neighbors (qABC).
Kiran and Findik [27] add the directional information into ABC and design a new search strategy, which selects the search
equation to generate a candidate solution according to the previous directional information (dABC). Since different search
equations have distinct advantages and perform differently on different problems or at different stages of the same prob-
lem, some methods with multiple search equations are consecutively proposed to enhance the comprehensive performance
of ABC, such as MEABC [43] , ABCVSS [28] and MuABC [13] . Moreover, Babaoglu [1] proposes a distribution-based solution
update rule instead of a differential update rule to address the stagnation behavior of ABC, which employs the mean and
standard deviation of the two selected food sources to generate a candidate solution (distABC).
2.2.2. Hybridization of ABC with auxiliary techniques
Like many EAs, the hybridization of ABC with other techniques could improve the performance of ABC. Banitalebi et al.
[2] propose an enhanced compact ABC by combining the Estimation of Distribution Algorithm (EDA) framework and ABC
(named EcABC), which not only inherits the memory-saving capability of EDA and the efficiency of ABC but also enhances
the exploitation ability of the original ABC. Moreover, Gao et al. use the orthogonal learning method [15] to improve the
performance of ABC. Recently, Li et al. [30] introduced a gene recombination operator into ABCs to combine the valuable in-
formation of different good solutions (GRABC). In addition, many local search techniques are introduced into ABC. Kang et al.