ORIGINAL ARTICLE
An improved artificial bee colony algorithm for minimal time cost
reduction
Jinling Cai
•
William Zhu
•
Haijun Ding
•
Fan Min
Received: 13 July 2013 / Accepted: 18 November 2013
Ó Springer-Verlag Berlin Heidelberg 2013
Abstract The artificial bee colony (ABC) is a popular
heuristic optimization algorithm. Although it has fewer
control parameters, it shows competitive performance
compared with other population-based algorithms. The
ABC algorithm is good at exploration, but poor at
exploitation. Recently, a global best-guided ABC (GABC)
algorithm, inspired by particle swarm optimization, has
been developed to tackle this issue. However, GABC
cannot be applied to binary optimization problems. In this
paper, we develop an improved ABC (IABC) algorithm
with a new food source update strategy. IABC employs
information about the global best solution as well as per-
sonal best solutions, thus enhancing the local search abil-
ities of the bees. The new algorithm is adjusted to solve the
binary optimization problem of minimal time cost reduc-
tion. We conduct a series of experiments on four UCI
datasets, and our results clearly indicate that our algorithm
outperforms the existing ABC algorithms, especially on the
medium-sized Mushroom dataset.
Keywords Artificial bee colony algorithm
Attribute reduction Testing time cost Waiting cost
1 Introduction
The artificial bee colony (ABC) is a new heuristic opti-
mization algorithm inspired by the intelligent foraging
behavior of honey bees. Despite having fewer control
parameters, it shows competitive performance compared
with other population-based algorithms. Karaboga [1]
compared the ABC algorithm with various optimization
algorithms on a large set of unconstrained numerical
functions. It was found that the performance of the ABC
algorithm was better than or similar to that of the other
techniques. Sundar [2] applied the ABC algorithm to solve
the quadratic minimum spanning tree problem (Q-MST),
and found that ABC obtained better quality Q-MST solu-
tions than genetic algorithms. Moreover, the ABC algo-
rithm has also been employed to train neural networks [3],
design infinite impulse response filters [4], reconfigure
distribution networks [5], plan wireless network paths [6],
and tune motor drives [7].
Results for the various applications mentioned above
demonstrate that the ABC algorithm has effective search
ability. However, its exploitation ability is poor. It is nec-
essary for a population-based optimization algorithm to
balance exploration and exploitation well [8]. In the ABC
algorithm, the employed bees and onlookers execute search
space exploitation, whereas exploration is left to scouts. In
the employed bee and onlooker stages, neighboring solu-
tions associated with every food source are randomly
constructed. Hence, there is no guarantee that these solu-
tions will be better than the previous one. Considering this
shortcoming, a global best-guided ABC (GABC) algorithm
[9] inspired by particle swarm optimization (PSO) [10, 11]
has been developed to improve the exploitation capability.
GABC incorporates information about the global best
solution into the neighboring solution search equation.
J. Cai W. Zhu (&) H. Ding
College of IOT Engineering, Hohai University,
Changzhou 213022, China
e-mail: williamfengzhu@gmail.com
W. Zhu F. Min
Lab of Granular Computing, Minnan Normal University,
Zhangzhou 363000, China
F. Min
Deparment of Computer Science, Southwest Petroleum
University, Chengdu 610500, China
123
Int. J. Mach. Learn. & Cyber.
DOI 10.1007/s13042-013-0219-8