Enhanced local search in artificial bee colony algorithm

International Association of Scientific Innovation and Research (IASIR)
(An Association Unifying the Sciences, Engineering, and Applied Research)
International Journal of Emerging Technologies in Computational
and Applied Sciences (IJETCAS)
www.iasir.net
IJETCAS 14-136; © 2014, IJETCAS All Rights Reserved Page 177
ISSN (Print): 2279-0047
ISSN (Online): 2279-0055
Enhanced Local Search in Artificial Bee Colony Algorithm
1
Sandeep Kumar, 2
Dr. Vivek Kumar Sharma, 3
Rajani Kumari
Faculty of Engineering and Technology
Jagannath University
Chaksu, Jaipur-303901
INDIA
Abstract: Artificial Bee Colony (ABC) algorithm is a Nature Inspired Algorithm (NIA) which based in intelligent
food foraging behaviour of honey bee swarm. ABC outperformed over other NIAs and other local search
heuristics when tested for benchmark functions as well as factual world problems but occasionally it shows
premature convergence and stagnation due to lack of balance between exploration and exploitation. This paper
establishes a local search mechanism that enhances exploration capability of ABC and avoids the dilemma of
stagnation. With help of recently introduces local search strategy it tries to balance intensification and
diversification of search space. The anticipated algorithm named as Enhanced local search in ABC (EnABC)
and tested over eleven benchmark functions. Results are evidence for its dominance over other competitive
algorithms.
Keywords: artificial bee colony algorithm; nature inspired algorithm; local search; memetic computing; swarm
intelligence
I. Introduction
Artificial bee colony Optimization algorithm is one of the trendy swam intelligence optimization algorithm
proposed by D. Karaboga [1]. These algorithms are used to come across a set of values for the non-aligned
variables that optimizes the value of one or more leaning variables. There are number of multivariable
optimization problems with haphazardly elevated dimensionality which cannot be solved by exact search
algorithms in evaluated time. So search algorithms capable of searching near-optimal or good solutions within up
to standard computation time are very useful in factual life. In last a small number of years, the community of
scientists and researchers has noticed the significance of a large number of nature-inspired metaheuristics and
hybrids of these nature-inspired optimization algorithms. Metaheuristics are identified as a general algorithmic
structure that can be employed to different optimization problems less number of modifications to adapt them for
a pinpointed problem. Metaheuristics are deliberated to extend the capabilities of heuristics by combining one or
more heuristic methods (referred to as procedures) using a higher-level methodologies (hence ‘meta’).
Metaheuristics are methods that provide direction to the search process. Hyper-heuristics are yet another
augmentation that focuses on heuristics that redefine their parameters (either online or offline) to enhance the
reliability of end result, or the effectiveness of the computation process. Hyperheuristics provide high-level
strategies that may employ machine learning and adapt their search behavior by modifying the relevance of the
sub-procedures or even which procedures are used (operating on the space of heuristics which in turn control
within the problem domain) [2]. Algorithms from the meadows of Computational intelligence, biologically
motivated intelligent computing, and Metaheuristics are applied to strenuous problems, to which more classical
approaches may not be applicable. Michalewicz and Fogel point towards that these problems are difficult [3] as:
they has large number of feasible solutions in the search space due to which they not able to exploit the best
results; The problem is so roundabout, that just to facilitate any conclusion at all, one need to use such librated
models of the problem that any result is essentially a waste of time; the evaluation function that describes the
quality of any proposed solution is noisy or varies with time, thereby requiring not just a single solution but an
entire series of solutions; the achievable solutions are so knowingly constrained that generating even on its own
unobjectionable answer is very hard-hitting job, let alone searching for an optimum solution; the human being
solving the problem is deficiently composed or assumes some psychological fencing that prevents them from
identifying precise solution.
Nature inspired algorithms are stimulated by some natural discernible fact, can be classified as per their source of
inspiration. Major categories of NIA are: Stochastic Algorithms, Evolutionary Algorithms, Swarm Algorithms,
Physical Algorithms, Probabilistic Algorithms, Immune Algorithms and Neural Algorithms [2].
Stochastic Algorithms are algorithms that focus on the introduction of randomness into heuristic methods.
Examples of stochastic algorithms are Random Search, Adaptive Random Search, Stochastic Hill Climbing,
Guided Local Search, Iterated Local Search, Variable Neighborhood Search, Greedy Randomized Adaptive, Tabu
Search, Reactive Tabu Search and Scatter Search [2].

Sandeep Kumar et al., International Journal of Emerging Technologies in Computational and Applied Sciences, 7(2), December 2013-
February, 2014, pp. 177-184
Evolutionary Algorithms are inspired by encroachment by means of innate assortment. Examples of evolutionary
algorithm are Genetic Algorithm, Genetic Programming, Learning Classifier System, Evolution Strategies,
Differential Evolution, Evolutionary Programming, Grammatical Evolution, Gene Expression Programming,
Non-dominated Sorting Genetic Algorithm, and Strength Pareto Evolutionary Algorithm [2].
Swarm Algorithms focus on strategies that exploit the properties of cooperative intelligence. Some examples of
swarm intelligence are Particle Swarm Optimization, Bees Algorithm, Ant System, Ant Colony System and
Bacterial Foraging Optimization Algorithm [2].
Physical Algorithms are inspired by corporal and societal systems like Simulated Annealing, Extremal
Optimization, Harmony Search, Cultural Algorithm, and Memetic Algorithm [2].
Probabilistic Algorithms are strategies that focus on methods that build models and educated guess distributions
in search domains. For example; Population-Based Incremental Learning, Univariate Marginal Distribution
Algorithm, Compact Genetic Algorithm, Bayesian Algorithm, Cross-Entropy Method [2].
Immune Algorithms are encouraged by the adaptive immune system of vertebrates like Clonal Selection
Algorithm, Negative Selection Algorithm, Artificial Immune Recognition System, Immune Network Algorithm
and Dendritic Cell Algorithm [2].
Neural Algorithms are inspired by the plasticity and learning qualities of the human nervous system. Some well
known neural algorithms are Perceptron, Back-propagation, Hopfield Network, Learning Vector Quantization
and Self-Organizing Map [2].
Real-world optimization problems and sweeping statements thereof can be strained from the largest part of
science, management, engineering, and information technology. Importantly, function optimization problems
have had a long tradition in the fields of Artificial Intelligence in motivating basic research into new problem
solving strategies, and for finding and verifying complete behavior against benchmark problem instances [1]. The
enhanced local search strategy proposed in this paper is based on one of the youngest constituent of NIA family
the Artificial Bee Colony (ABC) algorithm [1], which mimics the extra ordinary food foraging behavior of most
intelligent creepy-crawly that is honey bees swarm. ABC is a swarm-based intelligent stochastic optimization
algorithm that has shows considerable concert in solving continuous [4]-[6], combinatorial [7]-[10] and many
more complex optimization problems. Stochastic optimization algorithms are those that use unpredictability to
encourage non-deterministic characteristics, contrasted to entirely deterministic strategies. Most strategies from
the fields of biologically motivated Computation, Computational Intelligence and Metaheuristics may be
evaluated to belong the field of Stochastic Optimization [2].
The organization of rest of paper is as follow: In section 2, it initiate Artificial Bee Colony Algorithm in detailed
manner, one of the most up-to-date swarm based system introduces by D. Karaboga [1]. Next section discusses
some recent and important improvement and modifications in ABC. Section 4 introduces the proposed EnABC
algorithm. Section 5 contains experimental setup and results followed by conclusion and references.
II. Artificial Bee Colony Algorithm
The ABC algorithm mimics the extra ordinary food foraging behavior of the honey bee insects with three groups
of bees: employed bees, onlookers and scouts. A honey bee working to forage a food source (i.e. solution)
previously visited by itself and searching only around its locality is called an employed bee. Employed bees
accomplish waggle dance after coming back to the hive to barter the information of its food source to the rest of
the colony. A bee waiting in close proximity to the dance floor to designate any of the employed bees to follow is
known an onlooker. A bee randomly searching a search space for finding a food source is called a scout. For
every food source, there is only one employed bee and a number of follower bees. The scout bee, after finding
some food source better than some threshold value, what's more performs a waggle dance to share this
information. In the ABC algorithm implementation, half of the colony consists of employed bees and the other
half constitutes the onlookers. The number of food sources (i.e., solutions being exploited) is equal to the number
of employed bees in the colony. The employed bee whose food source (i.e. solution) is exhausted (i.e. the solution
has not been improved after several attempts) becomes a scout. The detailed algorithm is given below:
Step 1: Initialize population of N uniformly distributed individuals. Each individual xij is a food source and has D
attributes. D is the dimensionality of the problem. xij represent ith
solution in jth
dimension. Where j ∈ {1, 2, …,
D}
Step 2: Estimate the fitness of each individual solution using the following method,
if (solution value >= 0)
then
else
min max min[0,1]( )ij j j jx x rand x x  
( 1)2*
1
solution valueifit 

1(1 ( ))solutioni value
fit fabs 

February, 2014, pp. 177-184
Step 3: Each employed bee, placed at a food source that is different from others, search in the proximity of its
current position to find a better food source. For each employed bee, generate a new solution, vi around its current
position, xi using the following formula.
Here, k ∈ {1, 2,…, N} and j ∈ {1, 2, …, D} are randomly chosen indices. N is number of employed bees. ɸij is a
uniform random number from [-1, 1].
Step 4: Compute the fitness of both xi and vi. Apply greedy selection strategy to select better one of them.
Step 5: Calculate and normalize the probability values, Pi for each solution xi using the following formula.
Step 6: Assign each onlooker bee to a solution, xi at random with probability proportional to Pi.
Step 7: Generate new food positions (i.e. solutions), vi for each onlooker bee.
Step 8: Compute the fitness of each onlooker bee, xi and the new solution, vi. Apply greedy selection process to
keep the fitter one and abandon other.
Step 9: If a particular solution xi has not been improved over a predefined number of cycles, then select it for
rejection. Replace the solution by placing a scout bee at a food source generated evenly at random within the
search space using
for j = 1, 2,……,D
Step 10: Keep track of the best food sources (solution) found so far.
Step 11: Check termination criteria. If the best solution found is acceptable or reached the maximum iterations,
stop and return the best solution found so far. Otherwise go back to step 2 and repeat again.
III. Modifications in Artificial Bee colony Algorithm
Usually real world dispense some complex optimization problems that cannot be easily treated by available
mathematical optimization methods. If the user is not very aware about the exact solution of the problem in hand
then intelligence emerged from social behavior of social colony members may be used to solve these kinds of
problems. Honey bees are in the category of social insects. The foraging behavior of honey bees produces an
intelligent social behavior, called as swarm intelligence. This swarm intelligence is simulated and an intelligent
search algorithm namely, Artificial Bee Colony (ABC) algorithm is established by Karaboga in 2005 [1]. Since
its inception, a lot of research has been carried out to make ABC more and more efficient and to apply ABC for
different types of problems.
In order to get rid of the drawbacks of original ABC, researchers and scientists have modified ABC in many
ways. The potentials where ABC can be improved are fine tuning of ABC control parameters SN, ɸij and limit
(maximum cycle number). Hybridization of ABC with other population based probabilistic or deterministic
algorithms. New control parameters also introduced in different phases of ABC. D. Karaboga [1] has suggested
that the value of ɸij should be in the range of [-1, 1]. The value of limit (maximum cycle number) should be SN ×
D, where, SN is the number of solutions and D is the dimension of the problem. Wei-feng Gao et al. [11]
proposed an enhanced solution search equation in ABC, which is based on the fact that bee searches only around
the best solution of the previous iteration to increase the exploitation. A. Banharnsakun et al. [12] introduced a
new variant of ABC namely the best-so-far selection in artificial bee colony algorithm. To enhance the
exploitation and exploration processes, they propose to make three major changes by introducing the best-so-far
method, an adjustable search radius, and an objective-value-based comparison method in DE. J.C. Bansal et al.
[83] proposed balanced ABC; they introduced a new control parameter, Cognitive Learning Factor and also
modified range of ɸ in Artificial Bee Colony algorithm. Qingxian and Haijun proposed a modification in the
initialization scheme by making the initial group symmetrical, and the Boltzmann selection mechanism was
employed instead of roulette wheel selection for improving the convergence ability of the ABC algorithm [13].
In order to maximize the exploitation capacity of the onlooker stage, Tsai et al. introduced the Newtonian law of
universal gravitation in the onlooker phase of the basic ABC algorithm in which onlookers are selected based on a
roulette wheel (Interactive ABC, IABC) [14]. Baykasoglu et al. incorporated the ABC algorithm with shift
neighborhood searches and greedy randomized adaptive search heuristic and applied it to the generalized
assignment problem [15]. Furthermore, modified versions of the Artificial Bee Colony algorithm are introduced
and applied for efficiently solving real-parameter optimization problems by Bahriye Akay and Dervis Karaboga
[16]. To modify ABC behavior for constrained search space Mezura et al. [17] proposed four modifications
related with the selection strategy, the scout bee operator, and the equality and boundary constraints. Instead of
fitness proportional selection, tournament selection is performed to exploit employed bee food sources by
onlooker bees. Second, they employed dynamic tolerance for equality constraints. In 2010, Zhu and Kwong [18]
proposed an improved ABC algorithm called gbest-guided ABC (GABC) algorithm by incorporating the
information of global best (gbest) solution into the solution search equation to improve the exploitation. GABC is
inspired by PSO [19], which, in order to improve the exploitation capability, takes advantage of the information
of the global best (gbest) solution to guide the search by candidate solutions. J.C. Bansal et al. [24] introduced
( )ij ij ij kjijv x x x  
1
SN
i
i
ij
i
fit
P
fit




February, 2014, pp. 177-184
memetic search in ABC algorithm in order to balance exploitation and exploration. In 2010, Derelia and Das [20]
proposed a hybrid bee(s) algorithm for solving container loading problems. In the proposed algorithm, a bee(s)
algorithm is hybridized with the heuristic filling procedure for the solution of container loading problems. In
2010, Huang and Lin [21] proposed a new bee colony optimization algorithm with idle-time-based filtering
scheme and its application for open shop-scheduling problems. They categorized the foraging behaviors of bees
in two terms Forward Pass and Backward Pass. Forward Pass expresses the process of a forager bee leaving the
bee hive and flying towards a food source while Backward Pass denotes the process of a forager bee returning to
the bee hive and sharing the food source information with other forager bees (role change). In 2011, Nambiraj
Suguna et al. [22] proposed an independent rough set approach hybrid with artificial bee colony algorithm for
dimensionality reduction. In the proposed work, effects of the perturbation rate, the scaling factor (step size), and
the limit are investigated on real-parameter optimization. ABC algorithm hybridized with genetic algorithm to
balance exploration and exploitation of search space [25], [26]. In 2012, Bin Wu et al. [23] proposed
improvement of Global swarm optimization (GSO) hybrid with ABC and PSO. They use neighborhood solution
generation scheme of ABC and accept new solution only when it is better than previous one to improve GSO
performance. A detailed discussion on the performance and modification in ABC algorithm is available in
literature [27], [28].
IV. Enhenced Local Search in ABC(EnABC)
The proposed enhanced local search in ABC adds an additional step for local search inspired by MeABC [24].
The proposed algorithm introduces memetic search strategies as local search. The MeABC algorithm developed
by J. C. Bansal et al [24] influenced by Golden Section Search (GSS) [31]. In MeABC only the good particle of
the current swarm updates itself in its neighborhood. Original GSS method does not use any gradient
information of the function to finds the optima of a uni-modal continuous function. GSS processes the interval
[a = −1.2, b = 1.2] and initiates two intermediate points (f1, f2) with help of golden ratio (Ѱ = 0.618):
f1 = b − (b − a) × ψ and f2 = a + (b − a) × ψ
The proposed strategy use interval [a=-2, b=2] for GSS process in order to increase searching capability. The
detailed enhanced local search in ABC algorithm is given below:
Step 1: initialize the population of N uniformly distributed individuals. Each individual xij is a food source and
has D attributes. D is the dimensionality of the problem. Xij represent ith
solution in jth
dimension. Where j ∈ {1,
2, …, D}
Step 2: Estimate the fitness of each individual solution using the following method,
if (solution value >= 0)
then
else
Step 3: Each employed bee, placed at a food source that is different from others, search in the proximity of its
current position to find a better food source. For each employed bee, generate a new solution, vij around its
current position, xij using the following formula.
Here, k ∈ {1, 2,…, N} and j ∈ {1, 2, …, D} are randomly chosen indices. N is number of employed bees. ɸij is a
uniform random number from [-1, 1].
Step 4: Compute the fitness of both xij and vij. Apply greedy selection strategy to select better one of them.
Step 5: Calculate and normalize the probability values, Pij for each solution xi using the following formula.
Here ɸ =0.9.
Step 6: Assign each onlooker bee to a solution, xi at random with probability proportional to Pij.
Step 7: Generate new food positions, vij for each onlooker bee.
Step 8: Compute the fitness of each onlooker bee, xij and the new solution, vij. Apply greedy selection process to
keep the fitter one and abandon other.
Step 9: Memetic search phase inspired by GSS.
Repeat while termination criteria meet
Compute 1 ( ) ,f b b a     and 2 ( ) ,f a b a    
( 1)2*
1
solution valueifit 

1(1 ( ))solutioni value
fit fabs 
( )ij ij ij kjijv x x x  
1
_
(1 ) N
i
i i
ij
Maximum fitness
i
fit fit
P
fit
 

 


February, 2014, pp. 177-184
Calculate  1f f and  2f f
If  1f f <  2f f then
b = f2 and the solution lies in the range [a, b]
else
a = f1 and the solution lies in the range [a, b]
End of if
End of while
Step 10: If a particular solution xij has not been improved over a predefined number of cycles, then select it for
rejection. Replace the solution by placing a scout bee at a food source generated evenly at random within the
search space using
for j = 1, 2,……,D
Step 11: Keep track of the best food sources (solution) found so far.
Step 12: Check extinction criteria. If the best solution found is adequate or reached the maximum iterations, stop
and return the best solution found so far. Otherwise go back to step 2 and repeat all over again.
Table I Test Problems
Test
Problem
Objective Function Search
Range
Optimum
Value
D Acceptable
Error
Griewank
2
1 1 1
1
( ) ( ) cos 1
4000
DD i
ii i
x
f x x
i 
              
  [-600, 600] f(0) = 0 30 1.0E-05
Rastrigin
2
2
1
( ) 10cos(2 ) 10
D
i i
i
f x x x

   
  [-5.12,
5.12]
f(0) = 0 30 1.0E-05
Zakharov
2 41
3 1 1
1
2
( ) ( ) ( )
2 2
D Di
ii
i
D
i
ix ix
xf x
 

    [-5.12,
5.12]
f(0) = 0 30 1.0E-02
Salomon
Problem
2 2
4 1 1
( ) 1 cos(2 ) 0.1( )
D D
i ii i
f x x x
 
    [-100, 100] f(0) = 0 30 1.0E-01
Colville
function
2 2 2 2 2 2
5 2 1 1 4 3 3
2 2
2 4 2 4
( ) 100( ) (1 ) 90( ) (1 )
10.1[( 1) ( 1) ] 19.8( 1)( 1)
f x x x x x x x
x x x x
       
      
[-10, 10] f(1) = 0 4 1.0E-05
Kowalik
function
2
11 21 2
6 21
3 4
( )
( ) ( )i i
ii
i i
x b b x
f x a
b b x x

 
 
 [-5, 5]
f(0.1928,
0.1908,
0.1231,
0.1357) =
3.07E-04
4
1.0E-05
Shifted
Rosenbrock
1 2 2 2
7 11
1, 2 1 2
( ) (100( ) ( 1) ,
1, [ ,... ], [ , ,....... ]
D
i i i biasi
D D
f x z z z f
z x o x x x x o o o o


    
    
 [-100, 100] f(o)=fbias=390 10 1.0E-01
Six-hump
camel back
2 4 2 2 2
8 1 1 1 1 2 2 2
1
( ) (4 2.1 ) ( 4 4 )
3
f x x x x x x x x       [-5, 5]
f(-0.0898,
0.7126) = -
1.0316
2 1.0E-05
Easom’s
function
2 2
1 2( ( ) ( ) )
9 1 2( ) cos cos x x
f x x x e     
  [-10, 10] f(π, π) = -1 2 1.0E-13
McCormick 2
10 1 2 1 2 1 2
3 5
( ) sin( ) ( ) 1
2 2
f x x x x x x x      
1
2
1.5
4, 3
3,
x
x
 
  

f(-0.547, -
1.547) =-
1.9133
30 1.0E-04
Meyer and
Roth
Problem
5 21 3
11 1
1 2
( ) ( )
1
i
ii
i i
x x t
f x y
x t x v
 
  [-10, 10]
f(3.13,
15.16,0.78) =
0.4E-04
3 1.0E-03

February, 2014, pp. 177-184
V. Experimental Setting, Results and Discussion
A. Test problems under consideration
In order to analyze the performance of EnABC different global optimization problems (f1 to f11) are elected.
These are continuous optimization problems and have different degrees of complexity and multimodality. Test
problems are taken from [29], [30] with the associated offset values.
Table II Comparison of the Results of Test Problems
Test Function AlgorithmMeasure SD ME AFE SR
f1
ABC 9.10E-03 4.61E-03 81988.8 61
MeABC 1.41E-03 2.53E-04 39138.74 97
EnABC 1.03E-03 1.56E-04 79913.78 98
f2
ABC 1.43E+00 3.80E+00 99396.6 4
MeABC 1.63E-06 8.55E-06 42565.6 100
EnABC 2.74E-06 7.04E-06 41491.4 100
f3
ABC 1.49E+01 9.92E+01 100020 0
MeABC 7.87E-03 2.01E-02 99739.44 5
EnABC 5.25E-04 9.57E-03 116096.6 100
f4
ABC 2.30E-01 1.67E+00 100020.1 0
MeABC 3.47E-02 9.18E-01 20766.84 100
EnABC 3.10E-02 9.20E-01 26623.07 100
f5
ABC 1.36E-01 1.73E-01 98300.08 2
MeABC 2.65E-03 7.07E-03 35764.62 97
EnABC 1.40E-02 1.38E-02 109138.1 65
f6
ABC 8.04E-05 1.69E-04 86379.58 32
MeABC 5.37E-05 9.90E-05 57026.2 86
EnABC 6.40E-06 5.22E-06 28039.09 87
f7
ABC 9.79E+00 6.47E+00 97354.3 4
MeABC 1.71E+00 8.71E-01 74392.64 50
EnABC 4.12E+00 2.24E+00 178179.3 19
f8
ABC 1.06E-05 1.73E-06 78185.92 30
MeABC 4.24E-15 4.53E-15 5830.82 100
EnABC 1.41E-09 2.01E-10 95473.5 76
f9
ABC 4.02E-03 4.67E-03 100045.5 0
MeABC 1.39E-05 1.55E-05 54525.6 51
EnABC 1.37E-05 1.68E-05 52285.8 52
f10
ABC 2.46E-02 2.68E-02 100033 0
MeABC 1.33E-05 9.13E-06 52908.5 74
EnABC 7.65E-06 6.16E-06 80974.28 87
f11
ABC 4.16E-02 3.68E-02 100032.8 0
MeABC 1.20E-05 9.23E-05 39390.54 88
EnABC 4.93E-05 1.16E-04 126948.7 52
B. Experimental setting
To prove the efficiency of EnABC, it is compared with original ABC and MeABC. To test Enhanced local search
in ABC over considered problems, following experimental setting is adopted:
 Colony size N = 60,
 ɸij = rand[−1, 1],

February, 2014, pp. 177-184
 Number of food sources SN = N/2,
 limit = 1500,
 The stopping touchstone is either maximum number of function evaluations (which is set to be 200000)
is reached or the acceptable error (outlined in Table II) has been achieved,
 The number of simulations/run =100,
 Parameter settings for the original ABC algorithm and MeABC algorithm are same as EnABC.
C. Results Comparison
Mathematical results of EnABC with experimental setting as per subsection 5.B are outlined in Table II. Table
II show the relationship of results based on standard deviation (SD), mean error (ME), average function
evaluations (AFE) and success rate (SR) are reported. Table II shows that most of the time EnABC outperforms
in terms of efficiency (with less number of function evaluations) and reliability as compare to other considered
algorithms. The proposed algorithm all the time improves AFE and most of the time it also improve SD and
ME. It is due to randomness introduced during fitness calculation and probability calculation. Table III contains
summary of table II outcomes. In Table III, ‘+’ indicates that the EnABC is better than the considered
algorithms and ‘-’ indicates that the algorithm is not better or the difference is very small. The last row of Table
III, establishes the superiority of EnABC over MeABC and ABC.
Table III Summary of table II outcome
Function EnABC vs ABC EnABC vs MeABC
f1 + +
f2 + +
f3 + +
f4 + -
f5 + -
f6 + +
f7 + -
f8 + -
f9 + +
f10 + +
f11 + -
Total Number of + sign 11 6
VI. CONCLUSION
This paper, established a new phase in original ABC based on modified GSS process. Newly introduced step
added just after onlooker bee phase. Proposed algorithm also modifies search range of GSS process in order to
balance intensification and diversification of local search space. Further, the modified strategy is applied to solve
11 well-known standard benchmark functions. With the help of experiments over test problems, it is shown that
the insertion of the proposed local search strategy in the original ABC algorithm improves the steadfastness,
efficiency and accuracy as compare to their original version. Table II and III show that the proposed EnABC is
able to solve almost all the considered problems with fewer efforts.
REFERENCES
[1] D. Karaboga, “An idea based on honey bee swarm for numerical optimization”, Technical Report-TR06, Erciyes Univ.,
Computer Engg. Department, 2005.
[2] Brownlee, Clever algorithms: nature-inspired programming recipes. Jason Brownlee, 2011.
[3] Z. Michalewicz and D. B. Fogel. How to Solve It: Modern Heuristics. Springer, 2004.
[4] D. Karaboga, B. Basturk, Artificial Bee Colony (ABC) Optimization Algorithm for Solving Constrained Optimization Problems,
LNCS: Advances in Soft Computing: Foundations of Fuzzy Logic and Soft Computing, CA, Vol: 4529/2007, pp: 789-798,
Springer- Verlag, 2007, IFSA 2007.
[5] D. Karaboga, B. Basturk, “An artificial bee colony (ABC) algorithm for numeric function optimization”, in Proceedings of the
IEEE Swarm Intelligence Symposium 2006, Indianapolis, Indiana, USA, 12–14 May 2006.
[6] X. S. Yang, “Engineering Optimizations via Nature-Inspired Virtual Bee Algo.”, IWINAC2005, LNCS 3562, J. M. Yang and
J.R. Alvarez (Eds.), Springer-Verlag, Berlin Heidelberg, pp. 317-323, 2005.
[7] P. Lucic, D. Teodorovic, “Computing with Bees: Attacking Complex Transportation Engineering Problems”, in International
Journal on Artificial Intelligence Tools, vol. 12, no. 3, pp. 375-394, 2003.
[8] C. S. Chong, M. Y. H. Low, A. I. Sivakumar, K. L. Gay, “A Bee Colony Optimization Algorithm to Job Shop Scheduling”, Proc.
of the 37th Winter Simulation, California, pp. 1954-1961, 2006.
[9] G. Z. Markovic, D. Teodorovic, V. S. Acimovic-Raspopovic, “Routing and Wavelength Assignment in All-Optical Networks
Based on the Bee Colony Optimization”, AI Communications, v.20 no.4, p.273-285, December 2007.
[10] N. Quijano, K. M. Passino, “Honey Bee Social Foraging Algorithms for Resource Allocation Theory and Application”,
American Control Conference, New York City, USA, 2007.

February, 2014, pp. 177-184
[11] W. Gao and S. Liu. A modified artificial bee colony algorithm. Computers & Operations Research, 2011.
[12] A. Banharnsakun, T. Achalakul, and B. Sirinaovakul. The best-so-far selection in artificial bee colony algorithm. Applied Soft
Computing, 2010.
[13] D. Haijun F. Qingxian. Bee colony algorithm for the function optimization. Science Paper Online, August 2008.
[14] P.W. Tsai, J.S. Pan, B.Y. Liao, and S.C. Chu. Enhanced artificial bee colony optimization. International Journal of Innovative
Computing, Information and Control, 5(12), 2009.
[15] Baykasoglu, L. Ozbakir, and P. Tapkan. Artificial bee colony algorithm and its application to generalized assignment problem.
Swarm Intelligence: Focus on Ant and Particle Swarm Optimization, pages 113–144, 2007.
[16] B. Akay and D. Karaboga. A modified artificial bee colony algorithm for real-parameter optimization. Information Sciences,
2010.
[17] E. Mezura-Montes and O. Cetina-Dom´ınguez. Empirical analysis of a modified artificial bee colony for constrained numerical
optimization. Applied Mathematics and Computation, 2012.
[18] G. Zhu and S. Kwong. Gbest-guided artificial bee colony algorithm for numerical function optimization. Applied Mathematics
and Computation, 2010.
[19] J. Kennedy and R. C. Eberhart. Particle swarm optimization. In Proceedings IEEE int’l conf. on neural networks Vol. IV, pages
1942–1948, 1995.
[20] T. Derelia and G.S. Dasb. A hybrid’bee (s) algorithm’for solving container loading problems. Applied Soft Computing, 2010.
[21] Y.M. Huang and J.C. Lin. A new bee colony optimization algorithm with idle-time-based filtering scheme for open shop-
scheduling problems. Expert Systems with Applications, 2010.
[22] N. Suguna and K.G. Thanushkodi. An independent rough set approach hybrid with artificial bee colony algorithm for
dimensionality reduction. American Journal of Applied Sciences, 8, 2011.
[23] B. Wu, C. Qian, W. Ni, and S. Fan. The improvement of glowworm swarm optimization for continuous optimization problems.
Expert Systems with Applications, 2012.
[24] J.C. Bansal, H. Sharma, K.V. Arya and A. Nagar, “Memetic search in artificial bee colony algorithm.” Soft Computing (2013):
1-18.
[25] S. Kumar, V.K. Sharma, and R. Kumari. "A Novel Hybrid Crossover based Artificial Bee Colony Algorithm for Optimization
Problem." International Journal of Computer Applications 82, 2013.
[26] S. Pandey and S. Kumar "Enhanced Artificial Bee Colony Algorithm and It’s Application to Travelling Salesman Problem."
HCTL Open International Journal of Technology Innovations and Research, Volume 2, 2013:137-146.
[27] D. Karaboga, B. Gorkemli, C. Ozturk, and N. Karaboga. "A comprehensive survey: artificial bee colony (ABC) algorithm and
applications." Artificial Intelligence Review (2012): 1-37.
[28] J.C. Bansal, H. Sharma and S.S. Jadon "Artificial bee colony algorithm: a survey." International Journal of Advanced
Intelligence Paradigms 5.1 (2013): 123-159.
[29] M.M. Ali, C. Khompatraporn, and Z.B. Zabinsky. “A numerical evaluation of several stochastic algorithms on selected
continuous global optimization test problems.” Journal of Global Optimization, 31(4):635–672, 2005.
[30] P.N. Suganthan, N. Hansen, J.J. Liang, K. Deb, YP Chen, A. Auger, and S. Tiwari. “Problem definitions and evaluation criteria
for the CEC 2005 special session on real-parameter optimization.” In CEC 2005, 2005.
[31] J. Kiefer, “Sequential minimax search for a maximum.” In Proc. Amer. Math. Soc, volume 4, pages 502–506, 1953.
[32] S. Rahnamayan, H.R. Tizhoosh, and M.M.A. Salama. Opposition-based differential evolution. Evolutionary Computation, IEEE
Transactions on, 12(1):64–79, 2008.

Enhanced local search in artificial bee colony algorithm

More Related Content

What's hot (14)

Viewers also liked (14)

Similar to Enhanced local search in artificial bee colony algorithm (20)

More from Dr Sandeep Kumar Poonia (13)

Recently uploaded (20)

Enhanced local search in artificial bee colony algorithm