SlideShare uma empresa Scribd logo
1 de 9
Baixar para ler offline
International Journal of Information, Communication and Computing Technology
Jagan Institute of Management Studies, New Delhi
1
Research Scholar, Department of Computer Science & IT, Jagannath University, Jaipur
Email: 1
sandeep.kumar@jagannathuniversity.org
1
Assistant Professor, Department of Mathematics, Jagannath University, Jaipur
Email: 1
vivek.sharma@jagannathuniversity.org
1
Research Scholar, Department of Computer Science & IT, Jagannath University, Jaipur
Email: 1
rajanikpoonia@gmail.com
Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 20
Comparative study of Hybrids of Artificial Bee Colony Algorithm
1
Sandeep Kumar, 2
Dr. Vivek Kumar Sharma, 3
Rajani Kumari
ABSTRACT
Artificial bee colony (ABC) algorithm is a well known and one
of the latest swarm intelligence based techniques. This method
is a population based meta-heuristic algorithm used for
numerical optimization. It is based on the intelligent behavior
of honey bees. Artificial Bee Colony algorithm is one of the
most popular techniques that are used in optimization
problems. Artificial Bee Colony algorithm has some major
advantages over other heuristic methods. To utilize its good
feature a number of researchers combined ABC algorithm with
other methods, and generate some new hybrid methods. This
paper provides comparative analysis of hybrid differential
Artificial Bee Colony algorithm with hybrid ABC – SPSO,
Genetic algorithm and Independent rough set approach based
on some parameters like technique, dimension, methodology
etc.
KEYWORDS
Optimization techniques, Nature inspired techniques, Artificial
Bee Colony algorithm, DE, Genetic Algorithm, Rough Set,
PSO.
1. INTRODUCTION
Optimization problems have been solved by a number of
techniques. Now a day an alternative to the traditional methods
in operations research we use heuristic methods to solve
optimization problems.
The Nature Inspired methods are methods that are inspired by
natural and biological events, like immune system, foraging
behavior of ants and some other insects. Swarm intelligence is
one of the branches of nature inspired algorithms which are
used for function optimization. This branch is inspired by
intelligent behavior of insects [1]. Swarm intelligence an
algorithm leads to develop new meta-heuristics which can
utilize insect's problem solution abilities [2]. It has also been
investigated by a number of researchers that these meta
heuristic algorithms are able to provide very good solutions in
comparison to traditional algorithms. Optimization Techniques
is a unique reference source of methods for achieving the
optimal solution [3]. Optimization is stream computational
science in which we studies techniques for finding the best
possible solutions.
The organization of rest of paper is as follow: In section 2, we
introduce Artificial Bee Colony Algorithm, one of the newest
swarm based method introduces by D. Karaboga [4]. Next
section introduces hybrid optimization techniques. Section 4
introduces genetic algorithms. Next section describe hybrid of
ABC algorithm and Evolutionary Algorithm. Section 6 outlines
a hybrid of ABC and SPSO. Next section gives hybrid
algorithm of ABC and GA. Section 8 gives hybrid algorithm of
ABC and Rough Set approach. In next section we decided
some parameters to compare these four hybrid methods and
finally conclude the result in section 10.
2. ARTIFICIAL BEE COLONY ALGORITHM
The ABC algorithm is based on intelligent behavior of honey
bee swarm. This algorithm proposed by Dervis Karaboga in
2005 [4]. The major advantage of ABC over other swarm
intelligence methods is that in the ABC algorithm the possible
solutions indicate food sources (in this case flower); in spite of
individuals (honeybees swarm). In this method we divide bees
into two equal parts one is known as onlooker and other as
employee bees. Onlookers are applied to a food source
depending on the profitability. Onlookers compute a new
solution from its food source. When a food source is exhausted,
the bees employed on it converted into unemployed bees and
they have two options either they become a scout bee and
employed on another food source to exploit randomly or may
become an onlooker bees and waiting for information from
another bees about other food sources that are currently
exploited.
After fixed number of cycles (Maximum number of cycles that
is called as limit), if food source cannot be further improved, it
is rejected and replaced by randomly computed food source.
This is called exploration process and it takes place with help
of scout bees. It can be concluded that, employed and onlooker
bees are responsible for exploitation process, while scout bees
are engaged in process of exploration. Total number of
employed bees is equal to the number of food sources, each
employed bee represent a food source. Food source represents a
solution for the problem that need to be optimized. The basic
steps of the ABC algorithm are as given below [4], [5].
ALGORITHM 1: ABC ALGORITHM
 Compute the initial population of solutions
 Estimate the population
Comparative study of Hybrids of Artificial Bee Colony Algorithm
Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 21
o Compute new solutions for the employed
bees
o Apply the greedy selection mechanism
o Estimate the fitness and the probability
values
o Compute the new solutions for the onlookers
o Apply the greedy selection mechanism
o Determine the abandoned solution for the
scout bee, and change it with a latest
randomly generated solution
o Memorize the best solution achieved so far
 Repeat the above steps for pre decided limit
Artificial bee colony algorithm can be divided in four phases.
Each of the phases is outlined as follows:
2.1 Initialization of the swarm
The ABC algorithm has some parameters like: the number of
food sources, the number of trials (a threshold value after
which a food source is assumed to be discarded) and the
criteria for termination. In the original ABC, the number of
food sources is equal to the population of employed bees or
onlooker bees. Initially, a equally distributed initial swarm of
total food sources where each food source xi(i = 1, 2, ..., SN) is
a vector of D-dimension generated. Here D is taken as the
number of variables in the target problem and xi represent the
ith
food source in the swarm. Each food source is generated as
follows:
xij = xminj + rand[0,1] (xmaxj − xminj), (1)
here xminj and xmaxj are bounds of xi in jth
direction and rand [0,
1] is an evenly dispersed random number in the range [0, 1].
2.2 Employed bee phase
Second phase is employed bee phase; in this step modification
in the current solution (food source) is done by employed bees
as per the information gathered by individual experience and
the new fitness value. The position of ith
candidate updated
using equation 2.
(2)
here k ∈ {1, 2, ..., SN} and j ∈ {1, 2, ...,D} are arbitrarily taken
indices. k must be different from i. is a random number
between range [-1, 1].
2.3 Onlooker bees phase
In this phase, all the employed bees transfer information to
onlooker bees in the hive. Now they analyze this information
and select a solution with a probability probi related to its
fitness, which can be calculated using below mentioned
equation.
(3)
here fiti represent the fitness of the ith
solution.
2.4 Scout bees phase
If position of a food source not updates for some predecided
number of times then it is discarded. In this step, some bee
becomes scout whose food source has been rejected. The scout
bees change this food source by with help of equation 4.
xij = xminj + rand[0,1](xmaxj − xminj), for j∈{1,2,.,D} (4)
where xminj and xmaxj are bounds of xi in jth
direction.
The major reason behind use of ABC in case of separable
functions is that its modification step only modify exactly on
problem variable at a time after which the new solution is re-
computed. The performance of ABC algorithm has been
compared with some another optimization methods by many
researchers and they suggested modification in original ABC
algorithm in their way. One of those suggestions is a best-so-
far method for solution modification in the Artificial Bee
Colony algorithm was proposed in [6]. They replaced the
neighboring solutions-based approach with the best-so-far
technique to enhance the local searching power of the onlooker
bees. The searching method based on a dynamic adjustment of
search range depending on the iteration was introduced for
scout bees [6].
3. HYBRID OPTIMIZATION TECHNIQUES
Many techniques exist for combinatorial optimization.
Roughly, we will divide the sequence of accessible ways into
actual ways, that are in essence able to finding a best answer
and proving optimality however may have excessive
computation times, and heuristic ways together with
metaheuristics and evolutionary algorithms. Each world has
their specific properties, advantages, and disadvantages. We
work on the development of more practical hybrid optimization
systems exhibiting benefits of each world. In several
application areas, like network style, cutting and packing,
issues on graphs in general, scheduling, timetable management
and routing, such hybrids usually yield higher solutions in
shorter time. In distinction to classical heuristics, quality
guarantees or often even optimality-proofs will typically be
provided for obtained solutions [8].
In specific, we have a tendency to mix local-search-based
techniques like simulated annealing, tabu search, evolutionary
algorithms, and variable neighborhood search with clever
enumeration techniques using applied mathematics, like
cutting-plane algorithms, branch and prize, branch and cut and
branch and cut and prize. Beside the mix of metaheuristics with
International Journal of Information, Communication and Computing Technology (IJICCT)
Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 22
precise ways, we've additionally approached several issues by
mixtures of evolutionary algorithms and problem-specific
heuristics, native improvement and repair techniques. Such
mixtures are usually termed memetic algorithms [8].
The Nature Inspired methods have some advantages and
disadvantages as they are motivated by natural events. If these
methods are used individually they may inhibit some
weaknesses. These methods are complementary. To utilize the
advantages of various Nature inspired algorithm while avoiding
their drawbacks promotes advances in hybrid Nature inspired
algorithm. The idea behind motivation for hybridization of two
or more nature inspired computing algorithms is to enhance
rate of convergence, robustness, and reliability [7], [8].
The major goal behind development of new hybrids of ABC
algorithm is to improve rate of convergence. Hybrid Nature
inspired Computing methods [8] can be classified into different
categories according to the dimensions used, like application,
nature of problem, motivation for hybridization and
architecture of hybridization. In general, hybrid NIC methods
may be classified into completely different classes, based on
the measures used, like motivation for union and design of
union. We can divide them into „preprocessors and
postprocessors‟, „cooperators‟, and „embedded operators‟
supported by the link among the whole Nature impressed
formula concerned. Actually, a careful and comprehensive
analysis of the classification of the union would facilitate to
realize a deep understanding of the character impressed
formula and conjointly select the most effective combos for the
targeted improvement issues [8].
4. GENETIC ALGORITHM
Genetic Algorithm is traditional algorithm. This algorithm
proposed by Jhon Holland [14] in 1975. T A basic Genetic
Algorithm (GA) consists of 5 parts and major operators. First
of these part is a random number generator. Second is a fitness
analysis unit. Third part is genetic operators for reproduction,
forth is crossover phase and last is mutation operations. The
major algorithmic rules are outlined below:
 Initialize Population
 Repeat
o Analysis
o Reproduction
o Crossover
o Mutation
 Till the criteria met
The initial population needed at the beginning of the
algorithmic rule, could be a set of variety strings generated by
the random generator. Every string could be a illustration of an
answer to the improvement downside being addressed. Binary
strings square measure ordinarily used. Related to every string
could be a fitness price computed by the analysis unit. A fitness
value decides the quality of solution.
The aim of the genetic operators is to remodel this set of
strings into sets with higher fitness values. The reproduction
operator performs a simple selection process that is referred to
as seeded selection. Individual strings are derived from one set
to future per their fitness values, the higher the fitness worth,
the bigger the likelihood of a string being selected for future
generation. The crossover operator chooses pairs of strings
every which way and produces new pairs. The best crossover
operation is to chop the initial parent strings at a at random
selected purpose and to exchange their tails. The amount of
crossover operations is ruled by a crossover rate. The mutation
operator at random mutates or reverses the values of bits during
a string. The amount of mutation operations is set by a
mutation rate. A part of the formula consists of applying the
analysis, copy, crossover and mutation operations. A brand
new generation of solutions is made with every part of the
formula [13].
5. HYBRID DIFFERENTIAL ARTIFICIAL BEE
COLONY ALGORITHM
In this section we are going to discuss one of the powerful
heuristic methods, to solve non-linear, non-differentiable and
multi model optimization problems that is called as Differential
Evolution (DE). This simple and most powerful evolutionary
algorithm was given by Price and Storn. In last decade, DE
algorithm has become most popular in the field of machine
intelligence. Differential evolution combines mutation,
crossover and selection operators [9]. Differential Evolution an
evolutionary algorithm is a most powerful heuristic for solving
non-linear, non-differentiable and multi model optimization
problems [10].
The basic idea behind DE is a scheme for computing trail
parameter vectors. Mutation and crossover are used to compute
new vectors, and selection then after decides which of the
vectors have to be surviving for the next computation.
Classical evolutionary operators described by [9] can be stated
as follows:
 Mutation: Mutation is a process in which DE
generates a donor vector Vi_G corresponding to each
population individual (member) or target vector Xi_G
in the current generation, after initialization of
population.
 Crossover: Under this operation, the donor vector
exchanges its components with the target vector Xi_G
to form the trail vector.
 Selection: Selection determines which of the vectors
have to be surviving for the next generation. If the
new trail vector yields an equal or lower value of the
objective function, it alters the corresponding target
vector in the next generation; otherwise the target is
retained in the population.
Comparative study of Hybrids of Artificial Bee Colony Algorithm
Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 23
Hybridizing the Differential Evolution with heuristic
algorithms is expected to provide better convergence and
desired values. We proposed a new Hybrid Differential
Artificial Bee Colony (HDABC), which combines ABCA with
Differential Evolution algorithm. The goal of integrating DE
with ABC Algorithm is achieve better result, to combine their
advantages, and to remove their disadvantages [9].
The basic steps of HDABC Algorithm are as follows:
ALGORITHM 2: HDABC ALGORITHM
 Initialize all parameters
 Send the employed bees onto their corresponding food
sources and Compute their nectar amounts.
 Identify the position of the onlooker bees depending
upon the nectar amounts obtained by employed bees.
 Employ the scouts for exploiting new food sources.
 Memorize the best food sources identified so far.
 Select best (best fitness value) n differential vectors
from the population based on the fitness calculated to
generate initial population for Differential evolution
 DO
o FOR i = 1 to number of particles
o DO
 Mutation
 Crossover
 Selection
 END FOR
 If a termination criteria is not satisfied, go to second
step; otherwise stop the process and compute the best
solution is obtained so far [9].
The planned hybrid methodology additionally employs
identical constant quantity established that had been used by
the artificial Bee Colony algorithmic rule and Differential
Evolution. The sole amendment is that a number of the
numerical values are been accomplished than that of basic
parameters. During this hybrid theme, once the ABC completes
random improvement. Best ten food supplies are picked out
supported the fitness and these are the initial population for
Differential Evolution algorithmic rule. Then the population is
finely tuned by DE with support of the operators: Mutation,
crossover and choice. Because the best fitness value cannot be
obtained in single iteration that‟s why we set it as twenty
generations i.e., when each ABC iteration DE was allowed to
run supported by the user outlined generations. Combining
these 2 improvement techniques are expected to bring best
values with in lesser number of generations and additionally to
possess higher convergence speed [9].
6. A HYBRID ABC-SPSO ALGORITHM FOR
CONTINUOUS FUNCTION OPTIMIZATION
Particle Swarm optimization is based on swarm intelligence
algorithmic rule. This algorithmic rule simulates the movement
of a swarm of birds or a group of fish searching for food. PSO
is a metaheuristic because it makes few or no assumptions
concerning the matter being optimized and might search very
massive areas of candidate solutions. However, metaheuristics
like PSO don't guarantee for the best answer is ever found.
More specifically we can say that PSO doesn't use the gradient
of the matter being optimized, which implies PSO doesn't need
that the optimization problem be differentiable as is needed by
classic optimization strategies like gradient descent and quasi-
Newton strategies. PSO will thus even be used on optimization
issues that are partly irregular, noisy, modification over time,
etc. It has been wide applied to several benchmark functions
likewise as engineering applications with an excellent success.
The standard PSO (SPSO) algorithmic rule provides a choice
for a rotation of the random step in order to form it less
sensitive to rotations. Another advantage is that SPSO is
extremely prosperous for uni-modal optimization functions
[11].
ABC-SPSO is hybridizing two successful algorithms on the
components level in order to gain benefit from their respective
strengths [11]. This is hybridization of Particle Swarm
Optimization (PSO) with Artificial Bee Colony Algorithm
(ABC) [12]. The goal of integrating PSO with ABC is to
improve the personal bests of the particles. In hybrid algorithm
mix components from both ABC and SPSO in order to have an
algorithm that easily solve separable problems as ABC while
having a rotationally invariant behavior as SPSO at the same
time[11], [12].
For every particle 𝑖 in the swarm, the ABC algorithm update
equation is applied to its personal best pbest𝑖. This is done after
randomly selecting another particle 𝑘 and a random problem
variable 𝑗, hence pbest𝑖 is updated as follows:
𝑝𝑏𝑒𝑠𝑡𝑖𝑗 = 𝑝𝑏𝑒𝑠𝑡𝑖𝑗 + 𝜙𝑖𝑗 × (𝑝𝑏𝑒𝑠𝑡𝑖𝑗 − 𝑝𝑏𝑒𝑠𝑡 𝑘𝑗) (5)
The new pbest𝑖 value replaces the previous one if it has a better
fitness. These steps applied on the some benchmark functions
that are categorized into various classes like [11].
 Uni-modal functions
 Multimodal functions
 Hybrid functions
ALGORITHM 3: ABC-SPSO ALGORITHM
 Initialize the swarm population
 Evaluate the swarm
 Max_Iterations = 𝑀𝑎𝑥_𝐹𝑢𝑛𝑐𝑡𝑖𝑜𝑛_𝐸𝑣𝑎𝑙𝑢𝑎𝑡𝑖𝑜𝑛𝑠
𝑁𝑢𝑚_𝑃𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠
 Iteration_number=1
 WHILE 𝐼𝑡𝑒𝑟ation_𝑛𝑢𝑚𝑏𝑒𝑟 ≤ 𝑀𝑎𝑥_𝐼𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑠
 DO
o FOR( each particle 𝑖)
International Journal of Information, Communication and Computing Technology (IJICCT)
Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 24
o DO
 Modify v𝑖
 Modify x𝑖
 IF (pbest𝑖) ≤ (x𝑖) then
 pbest𝑖 = x𝑖
 end IF
o END FOR
o Update gbest
o FOR each particle 𝑖 do
 Choose a different random particle
 Choose a random problem variable
 Employ ABC update rule to pbest𝑖
 Modify pbesti and gbest
o END FOR
o Iteration_number = Iteration_number + 1
 END WHILE
 Return gbest
Above defined algorithm propose a hybridization of ABC and
SPSO. This is possible by incorporating an ABC component
into SPSO, which updates the personal best information of the
particles in each iteration using the ABC update method. Both
approaches are verified, one that updates all the particles using
the ABC component and one that updates the particles in
fitness proportionate approach. The proposed hybrid algorithm
able to retain the good SPSO performance in case of uni-modal
optimization problems while retaining the good ABC
performance on separable optimization problems as well.
However, the major disadvantage is that the ABC component is
only able to improve particles in a low percentage of the
computational power allocated to it [11].
7. HYBRID OF ARTIFICIAL BEE COLONY
ALGORITHM WITH GENETIC ALGORITHM FOR
NUMERICAL OPTIMIZATION
A newly proposed Hybrid Artificial Bee Colony (HABC)
algorithm, which combine canonical ABC with Genetic
Algorithm(GA) is use a real value single-point crossover
operator of Genetic Algorithm which is improve the canonical
ABC in solving complex optimization problems. This
algorithm deals with high dimension problems. This
optimization technique is used to find near-optimal solutions to
the complex numerical and engineering optimization problems.
Here only change in ABC algorithm is addition of crossover
phase among onlooker bees and scout bees [6], [7].
The crossover operator crosses two parent individuals to
produce new ones. Usually, the parent individuals are selected
with higher fitness. So, the good gene information will be
inherited and the individuals newly produced may be good
ones. A crossover point is generated randomly within the
dimension, and the two parent individuals exchange the values
of dimensions after the crossover point.
ALGORITHM 4: HABC
 Set Iteration_number =0;
 Initialize the food source positions;
 Compute the fitness value of food sources;
 WHILE
o FOR (every employed bee)
 Identify a new food source;
 Compute the fitness of the new food
source;
 Employ greedy selection method;
o END FOR.
 Calculate the probability P for each food source;
 FOR(each onlooker bee)
o Send onlooker bees to food sources
depending on Probability p;
o Produce a new food source;
o Compute the fitness of the new food source;
o Employ greedy selection method;
 END FOR
 Produce parent population applying tournament
selection;
 Select a certain amount of bad food sources;
 FOR(each selected food source)
o Select two parents randomly from the parent
population
o Produce two new food sources by crossing
the selected parents;
o Employ greedy selection for the selected
food source and the newly produced food
sources;
 END FOR
 IF(an employed bee becomes into a scout bee)
o Send the scout bee to a new randomly
produced food source;
 END IF
 Remember the best solution achieved so far;
o Iteration_number = Iteration_number +1;
 END WHILE
Now we can conclude that, the HABC algorithmic rule
employs a new management parameter to introduce a crossover
operator within the ABC algorithmic rule. With the new
operator, one can exchange additional data within the early
stage of the algorithmic rule, which reinforces the convergence
ability of the algorithmic rule. At the tip of the algorithmic rule,
the distinction between individuals‟ decreases and the agitation
of crossover operator also decreases. Population move towards
the improvement purpose.
Comparative study of Hybrids of Artificial Bee Colony Algorithm
Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 25
8. HYBRID OF ARTIFICIAL BEE COLONY
ALGORITHM WITH INDEPENDENT ROUGH SET
APPROACH FOR DIMENTIONALITY
REDUCTION
Rough set approach provides a mathematical mechanism that
can be used for machine learning, feature selection and
knowledge mining [8].
A new approach for dimensionality reduction proposed in [9].
This algorithm is hybrid of Artificial Bee Colony (ABC)
algorithm with independent Rough Set Approach. This
approach work in two steps:
 Identify the subset of attributes independently based
on decision attributes.
 Identify the final reduct [9].
This approach is used to find out the set of minimal attributes,
that is called „reducts‟ to classify objects without degradation
in quality of classification and generate minimal length
decision rules deep-rooted in a given information system. In
order to improve the performance of this method, an element of
pruning was introduced. Quick Reduct designate for a
dimensionality of n. For the worst-case dataset, evaluations of
the dependency function may be performed (n2
+n)/2 times. The
dependency of each attribute is estimated and it selects the best
candidate [9].
This hybrid algorithm is used to solve both constrained and
unconstrained type of optimization problems. This hybrid
algorithm uses the rough set theory in a number of real world
domains. The performance of this approach is analyzed with
five different medical datasets namely Dermatology, Cleveland
Heart, HIV, Lung Cancer and Wisconsin and compared with
six other reduct algorithms [9].The algorithm given by [9] is as
follow:
ALGORITHM 5: IQRBEE(Fc,Fd)
 Fc, the set of all conditional features;
 Fd, the set of decision features.
 Bundle the domain
 Identify the reduct for each class
 Build-up the core reduct and reduct sets
 Choose the initial parameter values for ABC
 Initialize the population (xi)
 Compute the objective and fitness value
 Find the optimum feature subset as global
 do
o Produce new feature subset(vi)
o Apply the greedy selection between xi and vi
o Calculate the fitness and probability values
o Generate the solutions for onlooker
o Employ the greedy selection for onlookers
o Decide the rejected solution and scouts
o Compute the cycle best feature subset
o Remember the best optimum feature subset
 Repeat up to max_number_of_cycles
The suggested algorithm can be used for feature reduction. Let
the bees choose the feature subsets arbitrarily and calculate
their fitness and realize the simplest one in every iteration. This
procedure is continual for variety of iterations to search out the
optimum set. As mentioned earlier, when selecting the core
reduct, with the remaining attributes at every cluster, the
utilized bee produces the feature subset in random. Consider a
domain that contains variety of distinctive decision values, then
an equivalent range of bees has been chosen because of the
population size. From this population half the bees are thought-
about as utilized bees and therefore the remaining half is
thought-about as onlooker bees. For every used bee, a random
subset from one reduct set is allotted. The random sets allotted
to all or any the bees are combined to make the feature subset
[9].
For instance, contemplate a database that contains ten numbers
of conditional attributes (c1,c2,…,c10) and three numbers of
call attributes with five hundred records. At the start the
records are bundled into three groups supported the choice
attribute then the reduct is applied for every cluster. As an
example, if we have a tendency to getting the reduct as:
R1 = {c1,c3,c4,c8}
R2 = {c3,c4,c9}
R3 = {c3,c4,c6,c7,c10}
From these reducts, the common attributes are chosen as base
reduct; during this example, Rc = {c3,c4} and also the
remaining attributes are off from every reduct:
R1 = {c1, C8}; R2 = { c9}; R3 = {c6,c7,c10}
Now three bees are employed to generate a reduct, by choosing
random subsets from these reducts and bundling them with the
base to find the optimum one. Like:
 Rc + Bee_1 = {c8} + Bee_2 = {c9} + Bee_3 =
{c7,c10}
 {c3,c4,c8,c9,c7,c10}
This new reduct is computed using the ABC algorithm. In the
second step of the algorithm, for every employed bee, That are
exactly half of the number of food sources, a new source is
generated as follow:
Vi,j = Xi,j +Ø(Xi,j – Xk,j) (6)
Where, Øij may be a uniformly distributed real random number
among the range [-1,1], k is the index of the answer chosen
indiscriminately from the colony (k = int (rand * N) + 1), j = 1,
International Journal of Information, Communication and Computing Technology (IJICCT)
Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 26
. . .,D and D is the dimension of the problem. When generating
vi, this new answer is compared to the existing solution Xi and
therefore the utilized bee exploits the higher supply. In the last
step of the algorithmic rule, a witness bee chooses a food
supply with the likelihood and produces a brand new supply in
designated food supply domain. As for utilized bee, the higher
supply is set to be exploited. The projected hybrid methodology
comes out with a far better optimal reduct than the others; that
shows its superior performance [9].
9. PARAMETER
In this paper, the following parameters are used for comparison
between hybridization of artificial bee colony algorithms.
 Hybrid
 Aim
 Techniques
 Methodology
 Dimension
 Developer
 Linear/Non linear
 Constrained/Unconstrained
 Known application area
The comparative analysis of the different hybrids of ABC
optimization technique with Genetic algorithm and
Independent rough set is given in form of table-1 at the end of
paper.
10. CONCLUSION
There are a large number of hybrids of NIC methods to solve
complex optimization problems. These hybrids can be
classified as per their applicability to the problem and nature of
problem. It may be possible that one problem can be solved by
many techniques. We can apply more than one optimization
techniques to solve the same problem. So in this situation, the
complexity plays an important role. It also matter through
which technique the best result is obtained whether that result
is optimum or not.
Now, the different techniques have different complexity. Any
problem must be solved with minimum complexity but also
provide the optimal solution. Here both ABC-SPSO and
HDABC techniques are able to solve constrained and
unconstrained problem, ABC-SPSO can be applicable in both
2D/3D where as HDABC is able to solve the problem of D-
dimensional space. ABC-SPSO and HDABC both can solve
non-linear problems. Different problems can find optimal
solution with different techniques. Hybrid Artificial Bee
Colony algorithm improve the canonical ABC in solving
complex optimization problem whereas Hybrid of Artificial
Bee Colony Algorithm with Independent Rough Set Approach
for Dimensionality Reduction can solve constrained/
unconstrained optimization problem and use mathematical
tools for both feature selection and knowledge discovery. This
paper these techniques are presented in an abstract way i.e. in
the form of table so that user can choose the technique
according to the problem.
REFERENCES
[1] L. A. Zadeh, “The roles of soft computing and fuzzy logic
in the conception, design and deployment of intelligent
system,” in Proceedings of the IEEE Asia Pacific
Conference on Circuits and Systems, Seoul, Korea,
November 1996, pp. 3-4.
[2] M. Subotic, M. Tuba, N. Stanarevic, “Parallelization of the
Artificial Bee Colony (ABC) Algorithm” Recent Advances
in Neural Networks, Fuzzy Systems & Evolutionary
Computing, ISSN: 1790-5109,2010, pp. 191-196.
[3] K. Price, “An Introduction to differential evolution,” New
Ideas in Optimization. D. Corne, M. Dorigo, and F. Glover
(Eds.), London, UK: McGraw Hill, 1999.
[4] D. Karaboga, “An Idea based on Bee Swarm for
Numerical Optimization”, Technical Report, TR-06,
Erciyes University Engineering Faculty, Computer
Engineering Department (2005).
[5] D. Karaboga, A. Basturk, “A powerful and efficient
algorithm for numerical function optimization: artificial
bee colony (ABC) algorithm”, J. of Global Optimization,
Volume 39, No. 0925-5001, 2007, pp. 459-471.
[6] A. Banharnsakun, T. Achalakul, B. Sirinaovakul, “The
best-so-far selection in Artificial Bee Colony algorithm”,
Applied Soft Computing 11 (2011), pp.2888–2901.
[7] L. N. de Castro, “Fundamentals of natural computing: An
overview,” Physics of Life Reviews, vol. 4, no. 1, pp. 1-36,
2007.
[8] X. Wang, X. Z. Gao, and S. J. Ovaska, “Hybrid Nature-
Inspired Computation Methods for Optimization” IEEE
Soft Computing Research, e-newslatter, December 2009,
issue#29
[9] A. Abraham, R. K. Jatoth, A. Rajasekhar, “Hybrid
Differential Artificial Bee Colony Algorithm” Journal of
computational and theoretical Nanoscience Vol.9,1-
9,2012.
[10]K. Price and R. Storn, Differential Evolution – a Simple
and Efficient Adaptive Scheme for Global Optimization
over Continuous Spaces, Technical Report, International
Computer Science Institute, Berkley (1995).
[11]Md. El-Abd, “A Hybrid ABC-SPSO Algorithm for
Continuous Function Optimization” in IEEE Symposium
on Swarm Intelligent System (SIS), 2011.
[12]J. Kennedy and R. C. Eberhart, “Particle swarm
optimization,” in Proceedings of IEEE International
Conference on Neural Networks, vol. 4. Piscataway, NJ:
IEEE Press, 1995, pp. 1942–1948.
[13]D. Karaboga, B. Akay, “A comparative study of Artificial
Bee Colony algorithm”. Applied Mathematics and
Computation 214 (2009), pp. 108–132
[14]J. J. Holland, Adaptation in Natural and Artificial Systems,
University of Michigan Press, 1975.
Comparative study of Hybrids of Artificial Bee Colony Algorithm
Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 27
Table 1. Comparative Analysis of Hybrid of ABC Algorithm
SN Parameter HABC ABC – PSO HDABC BEEIQR
1 Hybrid ABC and Genetic
Algorithm
ABC and SPSO ABC and DE ABC and RSAR
2 Aim Improve the
canonical ABC in
solving complex
optimization
problem.
Aimed at mixing
components from
both ABC and
SPSO in order to
have an algorithm
that easily solve
separable problem
as ABC while
having a
rotationally
invariant behavior
as SPSO at the
same time.
Optimization
process is applied
to each and every
individual in the
population
followed by
further
improvement
using Differential
Evolution search
Solve constrained/
unconstrained optimization
problem and use mathematical
tools for both feature selection
and knowledge discovery.
3 Technique Single point
crossover operator
of Genetic
Algorithm
Population based
search and
stochastic
optimization
method.
Mutation,
Crossover and
Selection operator
with Differential
Evolution.
Rough set theory in a number
of real world domains.
4 Methodolog
y
 Set i=0
 Initialize the
food source
positions.
 Evaluate the
nectar amount
(fitness).
 While the
termination
condition are
not met
 Execute
employed bees
phase
 Calculate the
probability P
for each food
source.
 Execute
onlooker bees
phase
 Execute
Crossover
phase
 Execute scout
 Initialize and
evaluate the
swarm
 Calculate Max
iterations
 For each
iteration
 For every
particle i
 Update vi and
xi
 If f(pbesti)<
f(xi)
 Pbesti = xi
 Update gbest
 For every
particle i
 Choose diff
random
particle and
random
problem
variable
 Apply ABC
 Initialization
While the
termination
criteria is not
met
 Apply ABC
algorithm
 Select best n
differential
vectors from
the
population
based on the
fitness
calculated to
generate
initial
population
for DE.
 For each
particle s
apply
Mutation
Crossover
Selection
 Stop
 Cluster domain
 Find reduct for each class
 Select initial parameter
for ABC
 Initialize population
 Calculate objective and
fitness value
 Find optimum feature
subset
 Repeat for maximum no
of cycles
 Produce initial solution
and apply greedy
selection
 Stop condition
International Journal of Information, Communication and Computing Technology (IJICCT)
Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 28
bees phase
 Memorize the
best solution
achieved so
far.
 i=i+1.
 Stop condition
update rule to
pbesti
 Update pbesti
and gbest
 Stop condition
condition
5 Dimension 2D/3D 2D/3D D- Dimension
space
-
6 Developer(s
)/
Proposer(s)
Xiaohui Yan,
Yunlong Zhu,
Wenping Zou
Mohammed El-Abd Ajith Abraham,
Ravi Kumar, A.
Rajasekhar
Nambiraj Suguna, Keppana
Gowder
7 Year 2011 2011 2012 2011
8 Linear/
Non Linear
Linear Non – linear Non - linear -
9 Constrained
/
Unconstrain
ed
Both Both Both Both
10 Known
application
area
Solve numerical
optimization
function
Function
optimization,
ANN, Fuzzy
control system
Real world
application:
Dynamic load
dispatch,
harmonic
estimation radar
tracking
Real world application:
Medical field
11 References [3] [4] [5] [6] [7] [8] [9] [10] [11]

Mais conteúdo relacionado

Mais procurados

Mais procurados (17)

An efficient and powerful advanced algorithm for solving real coded numerica...
An efficient and powerful advanced algorithm for solving real  coded numerica...An efficient and powerful advanced algorithm for solving real  coded numerica...
An efficient and powerful advanced algorithm for solving real coded numerica...
 
Cf34498502
Cf34498502Cf34498502
Cf34498502
 
Welcome to International Journal of Engineering Research and Development (IJERD)
Welcome to International Journal of Engineering Research and Development (IJERD)Welcome to International Journal of Engineering Research and Development (IJERD)
Welcome to International Journal of Engineering Research and Development (IJERD)
 
Evolutionary computing - soft computing
Evolutionary computing - soft computingEvolutionary computing - soft computing
Evolutionary computing - soft computing
 
Column store decision tree classification of unseen attribute set
Column store decision tree classification of unseen attribute setColumn store decision tree classification of unseen attribute set
Column store decision tree classification of unseen attribute set
 
Parallel evolutionary approach paper
Parallel evolutionary approach paperParallel evolutionary approach paper
Parallel evolutionary approach paper
 
Analysis of Parameter using Fuzzy Genetic Algorithm in E-learning System
Analysis of Parameter using Fuzzy Genetic Algorithm in E-learning SystemAnalysis of Parameter using Fuzzy Genetic Algorithm in E-learning System
Analysis of Parameter using Fuzzy Genetic Algorithm in E-learning System
 
Genetic Algorithms - Artificial Intelligence
Genetic Algorithms - Artificial IntelligenceGenetic Algorithms - Artificial Intelligence
Genetic Algorithms - Artificial Intelligence
 
Fuzzy Genetic Algorithm
Fuzzy Genetic AlgorithmFuzzy Genetic Algorithm
Fuzzy Genetic Algorithm
 
Introduction to the Genetic Algorithm
Introduction to the Genetic AlgorithmIntroduction to the Genetic Algorithm
Introduction to the Genetic Algorithm
 
A new CPXR Based Logistic Regression Method and Clinical Prognostic Modeling ...
A new CPXR Based Logistic Regression Method and Clinical Prognostic Modeling ...A new CPXR Based Logistic Regression Method and Clinical Prognostic Modeling ...
A new CPXR Based Logistic Regression Method and Clinical Prognostic Modeling ...
 
PORTFOLIO SELECTION BY THE MEANS OF CUCKOO OPTIMIZATION ALGORITHM
PORTFOLIO SELECTION BY THE MEANS OF CUCKOO OPTIMIZATION ALGORITHMPORTFOLIO SELECTION BY THE MEANS OF CUCKOO OPTIMIZATION ALGORITHM
PORTFOLIO SELECTION BY THE MEANS OF CUCKOO OPTIMIZATION ALGORITHM
 
THE STUDY OF CUCKOO OPTIMIZATION ALGORITHM FOR PRODUCTION PLANNING PROBLEM
THE STUDY OF CUCKOO OPTIMIZATION ALGORITHM FOR PRODUCTION PLANNING PROBLEMTHE STUDY OF CUCKOO OPTIMIZATION ALGORITHM FOR PRODUCTION PLANNING PROBLEM
THE STUDY OF CUCKOO OPTIMIZATION ALGORITHM FOR PRODUCTION PLANNING PROBLEM
 
Result analysis of mining fast frequent itemset using compacted data
Result analysis of mining fast frequent itemset using compacted dataResult analysis of mining fast frequent itemset using compacted data
Result analysis of mining fast frequent itemset using compacted data
 
Mining Irregular Association Rules based on Action & Non-action Type Data
Mining Irregular Association Rules based on Action & Non-action Type DataMining Irregular Association Rules based on Action & Non-action Type Data
Mining Irregular Association Rules based on Action & Non-action Type Data
 
RM 701 Genetic Algorithm and Fuzzy Logic lecture
RM 701 Genetic Algorithm and Fuzzy Logic lectureRM 701 Genetic Algorithm and Fuzzy Logic lecture
RM 701 Genetic Algorithm and Fuzzy Logic lecture
 
Ar03402580261
Ar03402580261Ar03402580261
Ar03402580261
 

Destaque

Memetic search in differential evolution algorithm
Memetic search in differential evolution algorithmMemetic search in differential evolution algorithm
Memetic search in differential evolution algorithm
Dr Sandeep Kumar Poonia
 

Destaque (14)

Splay tree
Splay treeSplay tree
Splay tree
 
Lecture27 linear programming
Lecture27 linear programmingLecture27 linear programming
Lecture27 linear programming
 
Memetic search in differential evolution algorithm
Memetic search in differential evolution algorithmMemetic search in differential evolution algorithm
Memetic search in differential evolution algorithm
 
Lecture25
Lecture25Lecture25
Lecture25
 
Splay trees by NIKHIL ARORA (www.internetnotes.in)
Splay trees by NIKHIL ARORA (www.internetnotes.in)Splay trees by NIKHIL ARORA (www.internetnotes.in)
Splay trees by NIKHIL ARORA (www.internetnotes.in)
 
Splay Tree
Splay TreeSplay Tree
Splay Tree
 
Lecture24
Lecture24Lecture24
Lecture24
 
Sunzip user tool for data reduction using huffman algorithm
Sunzip user tool for data reduction using huffman algorithmSunzip user tool for data reduction using huffman algorithm
Sunzip user tool for data reduction using huffman algorithm
 
2-3 Tree
2-3 Tree2-3 Tree
2-3 Tree
 
Multiplication of two 3 d sparse matrices using 1d arrays and linked lists
Multiplication of two 3 d sparse matrices using 1d arrays and linked listsMultiplication of two 3 d sparse matrices using 1d arrays and linked lists
Multiplication of two 3 d sparse matrices using 1d arrays and linked lists
 
Lecture26
Lecture26Lecture26
Lecture26
 
Soft computing
Soft computingSoft computing
Soft computing
 
AVL Tree
AVL TreeAVL Tree
AVL Tree
 
Lecture28 tsp
Lecture28 tspLecture28 tsp
Lecture28 tsp
 

Semelhante a Comparative study of_hybrids_of_artificial_bee_colony_algorithm

Solving np hard problem artificial bee colony algorithm
Solving np hard problem  artificial bee colony algorithmSolving np hard problem  artificial bee colony algorithm
Solving np hard problem artificial bee colony algorithm
IAEME Publication
 
Solving np hard problem artificial bee colony algorithm
Solving np hard problem  artificial bee colony algorithmSolving np hard problem  artificial bee colony algorithm
Solving np hard problem artificial bee colony algorithm
IAEME Publication
 
Solving np hard problem using artificial bee colony algorithm
Solving np hard problem using artificial bee colony algorithmSolving np hard problem using artificial bee colony algorithm
Solving np hard problem using artificial bee colony algorithm
IAEME Publication
 
Solving np hard problem using artificial bee colony algorithm
Solving np hard problem using artificial bee colony algorithmSolving np hard problem using artificial bee colony algorithm
Solving np hard problem using artificial bee colony algorithm
IAEME Publication
 
A Novel Approach of Image Ranking based on Enhanced Artificial Bee Colony Alg...
A Novel Approach of Image Ranking based on Enhanced Artificial Bee Colony Alg...A Novel Approach of Image Ranking based on Enhanced Artificial Bee Colony Alg...
A Novel Approach of Image Ranking based on Enhanced Artificial Bee Colony Alg...
ijsrd.com
 
Volume 14 issue 03 march 2014_ijcsms_march14_10_14_rahul
Volume 14  issue 03  march 2014_ijcsms_march14_10_14_rahulVolume 14  issue 03  march 2014_ijcsms_march14_10_14_rahul
Volume 14 issue 03 march 2014_ijcsms_march14_10_14_rahul
Deepak Agarwal
 

Semelhante a Comparative study of_hybrids_of_artificial_bee_colony_algorithm (20)

Solving np hard problem artificial bee colony algorithm
Solving np hard problem  artificial bee colony algorithmSolving np hard problem  artificial bee colony algorithm
Solving np hard problem artificial bee colony algorithm
 
Solving np hard problem artificial bee colony algorithm
Solving np hard problem  artificial bee colony algorithmSolving np hard problem  artificial bee colony algorithm
Solving np hard problem artificial bee colony algorithm
 
Solving np hard problem using artificial bee colony algorithm
Solving np hard problem using artificial bee colony algorithmSolving np hard problem using artificial bee colony algorithm
Solving np hard problem using artificial bee colony algorithm
 
Solving np hard problem using artificial bee colony algorithm
Solving np hard problem using artificial bee colony algorithmSolving np hard problem using artificial bee colony algorithm
Solving np hard problem using artificial bee colony algorithm
 
A Novel Approach of Image Ranking based on Enhanced Artificial Bee Colony Alg...
A Novel Approach of Image Ranking based on Enhanced Artificial Bee Colony Alg...A Novel Approach of Image Ranking based on Enhanced Artificial Bee Colony Alg...
A Novel Approach of Image Ranking based on Enhanced Artificial Bee Colony Alg...
 
Optimal k-means clustering using artificial bee colony algorithm with variab...
Optimal k-means clustering using artificial bee colony  algorithm with variab...Optimal k-means clustering using artificial bee colony  algorithm with variab...
Optimal k-means clustering using artificial bee colony algorithm with variab...
 
COMPARISON BETWEEN ARTIFICIAL BEE COLONY ALGORITHM, SHUFFLED FROG LEAPING ALG...
COMPARISON BETWEEN ARTIFICIAL BEE COLONY ALGORITHM, SHUFFLED FROG LEAPING ALG...COMPARISON BETWEEN ARTIFICIAL BEE COLONY ALGORITHM, SHUFFLED FROG LEAPING ALG...
COMPARISON BETWEEN ARTIFICIAL BEE COLONY ALGORITHM, SHUFFLED FROG LEAPING ALG...
 
Volume 14 issue 03 march 2014_ijcsms_march14_10_14_rahul
Volume 14  issue 03  march 2014_ijcsms_march14_10_14_rahulVolume 14  issue 03  march 2014_ijcsms_march14_10_14_rahul
Volume 14 issue 03 march 2014_ijcsms_march14_10_14_rahul
 
COMPARISON BETWEEN ARTIFICIAL BEE COLONY ALGORITHM, SHUFFLED FROG LEAPING ALG...
COMPARISON BETWEEN ARTIFICIAL BEE COLONY ALGORITHM, SHUFFLED FROG LEAPING ALG...COMPARISON BETWEEN ARTIFICIAL BEE COLONY ALGORITHM, SHUFFLED FROG LEAPING ALG...
COMPARISON BETWEEN ARTIFICIAL BEE COLONY ALGORITHM, SHUFFLED FROG LEAPING ALG...
 
An Hybrid Learning Approach using Particle Intelligence Dynamics and Bacteri...
An Hybrid Learning Approach using Particle Intelligence  Dynamics and Bacteri...An Hybrid Learning Approach using Particle Intelligence  Dynamics and Bacteri...
An Hybrid Learning Approach using Particle Intelligence Dynamics and Bacteri...
 
Evolutionary Computing Techniques for Software Effort Estimation
Evolutionary Computing Techniques for Software Effort EstimationEvolutionary Computing Techniques for Software Effort Estimation
Evolutionary Computing Techniques for Software Effort Estimation
 
EVOLUTIONARY COMPUTING TECHNIQUES FOR SOFTWARE EFFORT ESTIMATION
EVOLUTIONARY COMPUTING TECHNIQUES FOR SOFTWARE EFFORT ESTIMATIONEVOLUTIONARY COMPUTING TECHNIQUES FOR SOFTWARE EFFORT ESTIMATION
EVOLUTIONARY COMPUTING TECHNIQUES FOR SOFTWARE EFFORT ESTIMATION
 
EVOLUTIONARY COMPUTING TECHNIQUES FOR SOFTWARE EFFORT ESTIMATION
EVOLUTIONARY COMPUTING TECHNIQUES FOR SOFTWARE EFFORT ESTIMATIONEVOLUTIONARY COMPUTING TECHNIQUES FOR SOFTWARE EFFORT ESTIMATION
EVOLUTIONARY COMPUTING TECHNIQUES FOR SOFTWARE EFFORT ESTIMATION
 
Evaluation the efficiency of cuckoo
Evaluation the efficiency of cuckooEvaluation the efficiency of cuckoo
Evaluation the efficiency of cuckoo
 
MOCANAR: A Multi-Objective Cuckoo Search Algorithm for Numeric Association Ru...
MOCANAR: A Multi-Objective Cuckoo Search Algorithm for Numeric Association Ru...MOCANAR: A Multi-Objective Cuckoo Search Algorithm for Numeric Association Ru...
MOCANAR: A Multi-Objective Cuckoo Search Algorithm for Numeric Association Ru...
 
MOCANAR: A MULTI-OBJECTIVE CUCKOO SEARCH ALGORITHM FOR NUMERIC ASSOCIATION RU...
MOCANAR: A MULTI-OBJECTIVE CUCKOO SEARCH ALGORITHM FOR NUMERIC ASSOCIATION RU...MOCANAR: A MULTI-OBJECTIVE CUCKOO SEARCH ALGORITHM FOR NUMERIC ASSOCIATION RU...
MOCANAR: A MULTI-OBJECTIVE CUCKOO SEARCH ALGORITHM FOR NUMERIC ASSOCIATION RU...
 
ABC Algorithm.
ABC Algorithm.ABC Algorithm.
ABC Algorithm.
 
Advanced Optimization Techniques
Advanced Optimization TechniquesAdvanced Optimization Techniques
Advanced Optimization Techniques
 
Chicken Swarm as a Multi Step Algorithm for Global Optimization
Chicken Swarm as a Multi Step Algorithm for Global OptimizationChicken Swarm as a Multi Step Algorithm for Global Optimization
Chicken Swarm as a Multi Step Algorithm for Global Optimization
 
Comparison Between Pid Controllers for Gryphon Robot Optimized With Neuro-Fuz...
Comparison Between Pid Controllers for Gryphon Robot Optimized With Neuro-Fuz...Comparison Between Pid Controllers for Gryphon Robot Optimized With Neuro-Fuz...
Comparison Between Pid Controllers for Gryphon Robot Optimized With Neuro-Fuz...
 

Mais de Dr Sandeep Kumar Poonia

Performance evaluation of different routing protocols in wsn using different ...
Performance evaluation of different routing protocols in wsn using different ...Performance evaluation of different routing protocols in wsn using different ...
Performance evaluation of different routing protocols in wsn using different ...
Dr Sandeep Kumar Poonia
 

Mais de Dr Sandeep Kumar Poonia (13)

A new approach of program slicing
A new approach of program slicingA new approach of program slicing
A new approach of program slicing
 
Performance evaluation of different routing protocols in wsn using different ...
Performance evaluation of different routing protocols in wsn using different ...Performance evaluation of different routing protocols in wsn using different ...
Performance evaluation of different routing protocols in wsn using different ...
 
Database aggregation using metadata
Database aggregation using metadataDatabase aggregation using metadata
Database aggregation using metadata
 
Performance evaluation of diff routing protocols in wsn using difft network p...
Performance evaluation of diff routing protocols in wsn using difft network p...Performance evaluation of diff routing protocols in wsn using difft network p...
Performance evaluation of diff routing protocols in wsn using difft network p...
 
Lecture23
Lecture23Lecture23
Lecture23
 
Problems in parallel computations of tree functions
Problems in parallel computations of tree functionsProblems in parallel computations of tree functions
Problems in parallel computations of tree functions
 
Parallel Algorithms
Parallel AlgorithmsParallel Algorithms
Parallel Algorithms
 
Parallel Algorithms
Parallel AlgorithmsParallel Algorithms
Parallel Algorithms
 
Parallel Algorithms
Parallel AlgorithmsParallel Algorithms
Parallel Algorithms
 
Network flow problems
Network flow problemsNetwork flow problems
Network flow problems
 
Shortest Path in Graph
Shortest Path in GraphShortest Path in Graph
Shortest Path in Graph
 
Topological Sort
Topological SortTopological Sort
Topological Sort
 
Graph
GraphGraph
Graph
 

Último

Vishram Singh - Textbook of Anatomy Upper Limb and Thorax.. Volume 1 (1).pdf
Vishram Singh - Textbook of Anatomy  Upper Limb and Thorax.. Volume 1 (1).pdfVishram Singh - Textbook of Anatomy  Upper Limb and Thorax.. Volume 1 (1).pdf
Vishram Singh - Textbook of Anatomy Upper Limb and Thorax.. Volume 1 (1).pdf
ssuserdda66b
 
Activity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdfActivity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdf
ciinovamais
 
The basics of sentences session 3pptx.pptx
The basics of sentences session 3pptx.pptxThe basics of sentences session 3pptx.pptx
The basics of sentences session 3pptx.pptx
heathfieldcps1
 

Último (20)

Application orientated numerical on hev.ppt
Application orientated numerical on hev.pptApplication orientated numerical on hev.ppt
Application orientated numerical on hev.ppt
 
Vishram Singh - Textbook of Anatomy Upper Limb and Thorax.. Volume 1 (1).pdf
Vishram Singh - Textbook of Anatomy  Upper Limb and Thorax.. Volume 1 (1).pdfVishram Singh - Textbook of Anatomy  Upper Limb and Thorax.. Volume 1 (1).pdf
Vishram Singh - Textbook of Anatomy Upper Limb and Thorax.. Volume 1 (1).pdf
 
Single or Multiple melodic lines structure
Single or Multiple melodic lines structureSingle or Multiple melodic lines structure
Single or Multiple melodic lines structure
 
Kodo Millet PPT made by Ghanshyam bairwa college of Agriculture kumher bhara...
Kodo Millet  PPT made by Ghanshyam bairwa college of Agriculture kumher bhara...Kodo Millet  PPT made by Ghanshyam bairwa college of Agriculture kumher bhara...
Kodo Millet PPT made by Ghanshyam bairwa college of Agriculture kumher bhara...
 
Unit-V; Pricing (Pharma Marketing Management).pptx
Unit-V; Pricing (Pharma Marketing Management).pptxUnit-V; Pricing (Pharma Marketing Management).pptx
Unit-V; Pricing (Pharma Marketing Management).pptx
 
Accessible Digital Futures project (20/03/2024)
Accessible Digital Futures project (20/03/2024)Accessible Digital Futures project (20/03/2024)
Accessible Digital Futures project (20/03/2024)
 
How to Manage Global Discount in Odoo 17 POS
How to Manage Global Discount in Odoo 17 POSHow to Manage Global Discount in Odoo 17 POS
How to Manage Global Discount in Odoo 17 POS
 
How to Create and Manage Wizard in Odoo 17
How to Create and Manage Wizard in Odoo 17How to Create and Manage Wizard in Odoo 17
How to Create and Manage Wizard in Odoo 17
 
SOC 101 Demonstration of Learning Presentation
SOC 101 Demonstration of Learning PresentationSOC 101 Demonstration of Learning Presentation
SOC 101 Demonstration of Learning Presentation
 
Activity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdfActivity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdf
 
Fostering Friendships - Enhancing Social Bonds in the Classroom
Fostering Friendships - Enhancing Social Bonds  in the ClassroomFostering Friendships - Enhancing Social Bonds  in the Classroom
Fostering Friendships - Enhancing Social Bonds in the Classroom
 
Towards a code of practice for AI in AT.pptx
Towards a code of practice for AI in AT.pptxTowards a code of practice for AI in AT.pptx
Towards a code of practice for AI in AT.pptx
 
ICT role in 21st century education and it's challenges.
ICT role in 21st century education and it's challenges.ICT role in 21st century education and it's challenges.
ICT role in 21st century education and it's challenges.
 
Introduction to Nonprofit Accounting: The Basics
Introduction to Nonprofit Accounting: The BasicsIntroduction to Nonprofit Accounting: The Basics
Introduction to Nonprofit Accounting: The Basics
 
Key note speaker Neum_Admir Softic_ENG.pdf
Key note speaker Neum_Admir Softic_ENG.pdfKey note speaker Neum_Admir Softic_ENG.pdf
Key note speaker Neum_Admir Softic_ENG.pdf
 
Python Notes for mca i year students osmania university.docx
Python Notes for mca i year students osmania university.docxPython Notes for mca i year students osmania university.docx
Python Notes for mca i year students osmania university.docx
 
The basics of sentences session 3pptx.pptx
The basics of sentences session 3pptx.pptxThe basics of sentences session 3pptx.pptx
The basics of sentences session 3pptx.pptx
 
Sociology 101 Demonstration of Learning Exhibit
Sociology 101 Demonstration of Learning ExhibitSociology 101 Demonstration of Learning Exhibit
Sociology 101 Demonstration of Learning Exhibit
 
Understanding Accommodations and Modifications
Understanding  Accommodations and ModificationsUnderstanding  Accommodations and Modifications
Understanding Accommodations and Modifications
 
Dyslexia AI Workshop for Slideshare.pptx
Dyslexia AI Workshop for Slideshare.pptxDyslexia AI Workshop for Slideshare.pptx
Dyslexia AI Workshop for Slideshare.pptx
 

Comparative study of_hybrids_of_artificial_bee_colony_algorithm

  • 1. International Journal of Information, Communication and Computing Technology Jagan Institute of Management Studies, New Delhi 1 Research Scholar, Department of Computer Science & IT, Jagannath University, Jaipur Email: 1 sandeep.kumar@jagannathuniversity.org 1 Assistant Professor, Department of Mathematics, Jagannath University, Jaipur Email: 1 vivek.sharma@jagannathuniversity.org 1 Research Scholar, Department of Computer Science & IT, Jagannath University, Jaipur Email: 1 rajanikpoonia@gmail.com Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 20 Comparative study of Hybrids of Artificial Bee Colony Algorithm 1 Sandeep Kumar, 2 Dr. Vivek Kumar Sharma, 3 Rajani Kumari ABSTRACT Artificial bee colony (ABC) algorithm is a well known and one of the latest swarm intelligence based techniques. This method is a population based meta-heuristic algorithm used for numerical optimization. It is based on the intelligent behavior of honey bees. Artificial Bee Colony algorithm is one of the most popular techniques that are used in optimization problems. Artificial Bee Colony algorithm has some major advantages over other heuristic methods. To utilize its good feature a number of researchers combined ABC algorithm with other methods, and generate some new hybrid methods. This paper provides comparative analysis of hybrid differential Artificial Bee Colony algorithm with hybrid ABC – SPSO, Genetic algorithm and Independent rough set approach based on some parameters like technique, dimension, methodology etc. KEYWORDS Optimization techniques, Nature inspired techniques, Artificial Bee Colony algorithm, DE, Genetic Algorithm, Rough Set, PSO. 1. INTRODUCTION Optimization problems have been solved by a number of techniques. Now a day an alternative to the traditional methods in operations research we use heuristic methods to solve optimization problems. The Nature Inspired methods are methods that are inspired by natural and biological events, like immune system, foraging behavior of ants and some other insects. Swarm intelligence is one of the branches of nature inspired algorithms which are used for function optimization. This branch is inspired by intelligent behavior of insects [1]. Swarm intelligence an algorithm leads to develop new meta-heuristics which can utilize insect's problem solution abilities [2]. It has also been investigated by a number of researchers that these meta heuristic algorithms are able to provide very good solutions in comparison to traditional algorithms. Optimization Techniques is a unique reference source of methods for achieving the optimal solution [3]. Optimization is stream computational science in which we studies techniques for finding the best possible solutions. The organization of rest of paper is as follow: In section 2, we introduce Artificial Bee Colony Algorithm, one of the newest swarm based method introduces by D. Karaboga [4]. Next section introduces hybrid optimization techniques. Section 4 introduces genetic algorithms. Next section describe hybrid of ABC algorithm and Evolutionary Algorithm. Section 6 outlines a hybrid of ABC and SPSO. Next section gives hybrid algorithm of ABC and GA. Section 8 gives hybrid algorithm of ABC and Rough Set approach. In next section we decided some parameters to compare these four hybrid methods and finally conclude the result in section 10. 2. ARTIFICIAL BEE COLONY ALGORITHM The ABC algorithm is based on intelligent behavior of honey bee swarm. This algorithm proposed by Dervis Karaboga in 2005 [4]. The major advantage of ABC over other swarm intelligence methods is that in the ABC algorithm the possible solutions indicate food sources (in this case flower); in spite of individuals (honeybees swarm). In this method we divide bees into two equal parts one is known as onlooker and other as employee bees. Onlookers are applied to a food source depending on the profitability. Onlookers compute a new solution from its food source. When a food source is exhausted, the bees employed on it converted into unemployed bees and they have two options either they become a scout bee and employed on another food source to exploit randomly or may become an onlooker bees and waiting for information from another bees about other food sources that are currently exploited. After fixed number of cycles (Maximum number of cycles that is called as limit), if food source cannot be further improved, it is rejected and replaced by randomly computed food source. This is called exploration process and it takes place with help of scout bees. It can be concluded that, employed and onlooker bees are responsible for exploitation process, while scout bees are engaged in process of exploration. Total number of employed bees is equal to the number of food sources, each employed bee represent a food source. Food source represents a solution for the problem that need to be optimized. The basic steps of the ABC algorithm are as given below [4], [5]. ALGORITHM 1: ABC ALGORITHM  Compute the initial population of solutions  Estimate the population
  • 2. Comparative study of Hybrids of Artificial Bee Colony Algorithm Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 21 o Compute new solutions for the employed bees o Apply the greedy selection mechanism o Estimate the fitness and the probability values o Compute the new solutions for the onlookers o Apply the greedy selection mechanism o Determine the abandoned solution for the scout bee, and change it with a latest randomly generated solution o Memorize the best solution achieved so far  Repeat the above steps for pre decided limit Artificial bee colony algorithm can be divided in four phases. Each of the phases is outlined as follows: 2.1 Initialization of the swarm The ABC algorithm has some parameters like: the number of food sources, the number of trials (a threshold value after which a food source is assumed to be discarded) and the criteria for termination. In the original ABC, the number of food sources is equal to the population of employed bees or onlooker bees. Initially, a equally distributed initial swarm of total food sources where each food source xi(i = 1, 2, ..., SN) is a vector of D-dimension generated. Here D is taken as the number of variables in the target problem and xi represent the ith food source in the swarm. Each food source is generated as follows: xij = xminj + rand[0,1] (xmaxj − xminj), (1) here xminj and xmaxj are bounds of xi in jth direction and rand [0, 1] is an evenly dispersed random number in the range [0, 1]. 2.2 Employed bee phase Second phase is employed bee phase; in this step modification in the current solution (food source) is done by employed bees as per the information gathered by individual experience and the new fitness value. The position of ith candidate updated using equation 2. (2) here k ∈ {1, 2, ..., SN} and j ∈ {1, 2, ...,D} are arbitrarily taken indices. k must be different from i. is a random number between range [-1, 1]. 2.3 Onlooker bees phase In this phase, all the employed bees transfer information to onlooker bees in the hive. Now they analyze this information and select a solution with a probability probi related to its fitness, which can be calculated using below mentioned equation. (3) here fiti represent the fitness of the ith solution. 2.4 Scout bees phase If position of a food source not updates for some predecided number of times then it is discarded. In this step, some bee becomes scout whose food source has been rejected. The scout bees change this food source by with help of equation 4. xij = xminj + rand[0,1](xmaxj − xminj), for j∈{1,2,.,D} (4) where xminj and xmaxj are bounds of xi in jth direction. The major reason behind use of ABC in case of separable functions is that its modification step only modify exactly on problem variable at a time after which the new solution is re- computed. The performance of ABC algorithm has been compared with some another optimization methods by many researchers and they suggested modification in original ABC algorithm in their way. One of those suggestions is a best-so- far method for solution modification in the Artificial Bee Colony algorithm was proposed in [6]. They replaced the neighboring solutions-based approach with the best-so-far technique to enhance the local searching power of the onlooker bees. The searching method based on a dynamic adjustment of search range depending on the iteration was introduced for scout bees [6]. 3. HYBRID OPTIMIZATION TECHNIQUES Many techniques exist for combinatorial optimization. Roughly, we will divide the sequence of accessible ways into actual ways, that are in essence able to finding a best answer and proving optimality however may have excessive computation times, and heuristic ways together with metaheuristics and evolutionary algorithms. Each world has their specific properties, advantages, and disadvantages. We work on the development of more practical hybrid optimization systems exhibiting benefits of each world. In several application areas, like network style, cutting and packing, issues on graphs in general, scheduling, timetable management and routing, such hybrids usually yield higher solutions in shorter time. In distinction to classical heuristics, quality guarantees or often even optimality-proofs will typically be provided for obtained solutions [8]. In specific, we have a tendency to mix local-search-based techniques like simulated annealing, tabu search, evolutionary algorithms, and variable neighborhood search with clever enumeration techniques using applied mathematics, like cutting-plane algorithms, branch and prize, branch and cut and branch and cut and prize. Beside the mix of metaheuristics with
  • 3. International Journal of Information, Communication and Computing Technology (IJICCT) Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 22 precise ways, we've additionally approached several issues by mixtures of evolutionary algorithms and problem-specific heuristics, native improvement and repair techniques. Such mixtures are usually termed memetic algorithms [8]. The Nature Inspired methods have some advantages and disadvantages as they are motivated by natural events. If these methods are used individually they may inhibit some weaknesses. These methods are complementary. To utilize the advantages of various Nature inspired algorithm while avoiding their drawbacks promotes advances in hybrid Nature inspired algorithm. The idea behind motivation for hybridization of two or more nature inspired computing algorithms is to enhance rate of convergence, robustness, and reliability [7], [8]. The major goal behind development of new hybrids of ABC algorithm is to improve rate of convergence. Hybrid Nature inspired Computing methods [8] can be classified into different categories according to the dimensions used, like application, nature of problem, motivation for hybridization and architecture of hybridization. In general, hybrid NIC methods may be classified into completely different classes, based on the measures used, like motivation for union and design of union. We can divide them into „preprocessors and postprocessors‟, „cooperators‟, and „embedded operators‟ supported by the link among the whole Nature impressed formula concerned. Actually, a careful and comprehensive analysis of the classification of the union would facilitate to realize a deep understanding of the character impressed formula and conjointly select the most effective combos for the targeted improvement issues [8]. 4. GENETIC ALGORITHM Genetic Algorithm is traditional algorithm. This algorithm proposed by Jhon Holland [14] in 1975. T A basic Genetic Algorithm (GA) consists of 5 parts and major operators. First of these part is a random number generator. Second is a fitness analysis unit. Third part is genetic operators for reproduction, forth is crossover phase and last is mutation operations. The major algorithmic rules are outlined below:  Initialize Population  Repeat o Analysis o Reproduction o Crossover o Mutation  Till the criteria met The initial population needed at the beginning of the algorithmic rule, could be a set of variety strings generated by the random generator. Every string could be a illustration of an answer to the improvement downside being addressed. Binary strings square measure ordinarily used. Related to every string could be a fitness price computed by the analysis unit. A fitness value decides the quality of solution. The aim of the genetic operators is to remodel this set of strings into sets with higher fitness values. The reproduction operator performs a simple selection process that is referred to as seeded selection. Individual strings are derived from one set to future per their fitness values, the higher the fitness worth, the bigger the likelihood of a string being selected for future generation. The crossover operator chooses pairs of strings every which way and produces new pairs. The best crossover operation is to chop the initial parent strings at a at random selected purpose and to exchange their tails. The amount of crossover operations is ruled by a crossover rate. The mutation operator at random mutates or reverses the values of bits during a string. The amount of mutation operations is set by a mutation rate. A part of the formula consists of applying the analysis, copy, crossover and mutation operations. A brand new generation of solutions is made with every part of the formula [13]. 5. HYBRID DIFFERENTIAL ARTIFICIAL BEE COLONY ALGORITHM In this section we are going to discuss one of the powerful heuristic methods, to solve non-linear, non-differentiable and multi model optimization problems that is called as Differential Evolution (DE). This simple and most powerful evolutionary algorithm was given by Price and Storn. In last decade, DE algorithm has become most popular in the field of machine intelligence. Differential evolution combines mutation, crossover and selection operators [9]. Differential Evolution an evolutionary algorithm is a most powerful heuristic for solving non-linear, non-differentiable and multi model optimization problems [10]. The basic idea behind DE is a scheme for computing trail parameter vectors. Mutation and crossover are used to compute new vectors, and selection then after decides which of the vectors have to be surviving for the next computation. Classical evolutionary operators described by [9] can be stated as follows:  Mutation: Mutation is a process in which DE generates a donor vector Vi_G corresponding to each population individual (member) or target vector Xi_G in the current generation, after initialization of population.  Crossover: Under this operation, the donor vector exchanges its components with the target vector Xi_G to form the trail vector.  Selection: Selection determines which of the vectors have to be surviving for the next generation. If the new trail vector yields an equal or lower value of the objective function, it alters the corresponding target vector in the next generation; otherwise the target is retained in the population.
  • 4. Comparative study of Hybrids of Artificial Bee Colony Algorithm Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 23 Hybridizing the Differential Evolution with heuristic algorithms is expected to provide better convergence and desired values. We proposed a new Hybrid Differential Artificial Bee Colony (HDABC), which combines ABCA with Differential Evolution algorithm. The goal of integrating DE with ABC Algorithm is achieve better result, to combine their advantages, and to remove their disadvantages [9]. The basic steps of HDABC Algorithm are as follows: ALGORITHM 2: HDABC ALGORITHM  Initialize all parameters  Send the employed bees onto their corresponding food sources and Compute their nectar amounts.  Identify the position of the onlooker bees depending upon the nectar amounts obtained by employed bees.  Employ the scouts for exploiting new food sources.  Memorize the best food sources identified so far.  Select best (best fitness value) n differential vectors from the population based on the fitness calculated to generate initial population for Differential evolution  DO o FOR i = 1 to number of particles o DO  Mutation  Crossover  Selection  END FOR  If a termination criteria is not satisfied, go to second step; otherwise stop the process and compute the best solution is obtained so far [9]. The planned hybrid methodology additionally employs identical constant quantity established that had been used by the artificial Bee Colony algorithmic rule and Differential Evolution. The sole amendment is that a number of the numerical values are been accomplished than that of basic parameters. During this hybrid theme, once the ABC completes random improvement. Best ten food supplies are picked out supported the fitness and these are the initial population for Differential Evolution algorithmic rule. Then the population is finely tuned by DE with support of the operators: Mutation, crossover and choice. Because the best fitness value cannot be obtained in single iteration that‟s why we set it as twenty generations i.e., when each ABC iteration DE was allowed to run supported by the user outlined generations. Combining these 2 improvement techniques are expected to bring best values with in lesser number of generations and additionally to possess higher convergence speed [9]. 6. A HYBRID ABC-SPSO ALGORITHM FOR CONTINUOUS FUNCTION OPTIMIZATION Particle Swarm optimization is based on swarm intelligence algorithmic rule. This algorithmic rule simulates the movement of a swarm of birds or a group of fish searching for food. PSO is a metaheuristic because it makes few or no assumptions concerning the matter being optimized and might search very massive areas of candidate solutions. However, metaheuristics like PSO don't guarantee for the best answer is ever found. More specifically we can say that PSO doesn't use the gradient of the matter being optimized, which implies PSO doesn't need that the optimization problem be differentiable as is needed by classic optimization strategies like gradient descent and quasi- Newton strategies. PSO will thus even be used on optimization issues that are partly irregular, noisy, modification over time, etc. It has been wide applied to several benchmark functions likewise as engineering applications with an excellent success. The standard PSO (SPSO) algorithmic rule provides a choice for a rotation of the random step in order to form it less sensitive to rotations. Another advantage is that SPSO is extremely prosperous for uni-modal optimization functions [11]. ABC-SPSO is hybridizing two successful algorithms on the components level in order to gain benefit from their respective strengths [11]. This is hybridization of Particle Swarm Optimization (PSO) with Artificial Bee Colony Algorithm (ABC) [12]. The goal of integrating PSO with ABC is to improve the personal bests of the particles. In hybrid algorithm mix components from both ABC and SPSO in order to have an algorithm that easily solve separable problems as ABC while having a rotationally invariant behavior as SPSO at the same time[11], [12]. For every particle 𝑖 in the swarm, the ABC algorithm update equation is applied to its personal best pbest𝑖. This is done after randomly selecting another particle 𝑘 and a random problem variable 𝑗, hence pbest𝑖 is updated as follows: 𝑝𝑏𝑒𝑠𝑡𝑖𝑗 = 𝑝𝑏𝑒𝑠𝑡𝑖𝑗 + 𝜙𝑖𝑗 × (𝑝𝑏𝑒𝑠𝑡𝑖𝑗 − 𝑝𝑏𝑒𝑠𝑡 𝑘𝑗) (5) The new pbest𝑖 value replaces the previous one if it has a better fitness. These steps applied on the some benchmark functions that are categorized into various classes like [11].  Uni-modal functions  Multimodal functions  Hybrid functions ALGORITHM 3: ABC-SPSO ALGORITHM  Initialize the swarm population  Evaluate the swarm  Max_Iterations = 𝑀𝑎𝑥_𝐹𝑢𝑛𝑐𝑡𝑖𝑜𝑛_𝐸𝑣𝑎𝑙𝑢𝑎𝑡𝑖𝑜𝑛𝑠 𝑁𝑢𝑚_𝑃𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠  Iteration_number=1  WHILE 𝐼𝑡𝑒𝑟ation_𝑛𝑢𝑚𝑏𝑒𝑟 ≤ 𝑀𝑎𝑥_𝐼𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑠  DO o FOR( each particle 𝑖)
  • 5. International Journal of Information, Communication and Computing Technology (IJICCT) Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 24 o DO  Modify v𝑖  Modify x𝑖  IF (pbest𝑖) ≤ (x𝑖) then  pbest𝑖 = x𝑖  end IF o END FOR o Update gbest o FOR each particle 𝑖 do  Choose a different random particle  Choose a random problem variable  Employ ABC update rule to pbest𝑖  Modify pbesti and gbest o END FOR o Iteration_number = Iteration_number + 1  END WHILE  Return gbest Above defined algorithm propose a hybridization of ABC and SPSO. This is possible by incorporating an ABC component into SPSO, which updates the personal best information of the particles in each iteration using the ABC update method. Both approaches are verified, one that updates all the particles using the ABC component and one that updates the particles in fitness proportionate approach. The proposed hybrid algorithm able to retain the good SPSO performance in case of uni-modal optimization problems while retaining the good ABC performance on separable optimization problems as well. However, the major disadvantage is that the ABC component is only able to improve particles in a low percentage of the computational power allocated to it [11]. 7. HYBRID OF ARTIFICIAL BEE COLONY ALGORITHM WITH GENETIC ALGORITHM FOR NUMERICAL OPTIMIZATION A newly proposed Hybrid Artificial Bee Colony (HABC) algorithm, which combine canonical ABC with Genetic Algorithm(GA) is use a real value single-point crossover operator of Genetic Algorithm which is improve the canonical ABC in solving complex optimization problems. This algorithm deals with high dimension problems. This optimization technique is used to find near-optimal solutions to the complex numerical and engineering optimization problems. Here only change in ABC algorithm is addition of crossover phase among onlooker bees and scout bees [6], [7]. The crossover operator crosses two parent individuals to produce new ones. Usually, the parent individuals are selected with higher fitness. So, the good gene information will be inherited and the individuals newly produced may be good ones. A crossover point is generated randomly within the dimension, and the two parent individuals exchange the values of dimensions after the crossover point. ALGORITHM 4: HABC  Set Iteration_number =0;  Initialize the food source positions;  Compute the fitness value of food sources;  WHILE o FOR (every employed bee)  Identify a new food source;  Compute the fitness of the new food source;  Employ greedy selection method; o END FOR.  Calculate the probability P for each food source;  FOR(each onlooker bee) o Send onlooker bees to food sources depending on Probability p; o Produce a new food source; o Compute the fitness of the new food source; o Employ greedy selection method;  END FOR  Produce parent population applying tournament selection;  Select a certain amount of bad food sources;  FOR(each selected food source) o Select two parents randomly from the parent population o Produce two new food sources by crossing the selected parents; o Employ greedy selection for the selected food source and the newly produced food sources;  END FOR  IF(an employed bee becomes into a scout bee) o Send the scout bee to a new randomly produced food source;  END IF  Remember the best solution achieved so far; o Iteration_number = Iteration_number +1;  END WHILE Now we can conclude that, the HABC algorithmic rule employs a new management parameter to introduce a crossover operator within the ABC algorithmic rule. With the new operator, one can exchange additional data within the early stage of the algorithmic rule, which reinforces the convergence ability of the algorithmic rule. At the tip of the algorithmic rule, the distinction between individuals‟ decreases and the agitation of crossover operator also decreases. Population move towards the improvement purpose.
  • 6. Comparative study of Hybrids of Artificial Bee Colony Algorithm Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 25 8. HYBRID OF ARTIFICIAL BEE COLONY ALGORITHM WITH INDEPENDENT ROUGH SET APPROACH FOR DIMENTIONALITY REDUCTION Rough set approach provides a mathematical mechanism that can be used for machine learning, feature selection and knowledge mining [8]. A new approach for dimensionality reduction proposed in [9]. This algorithm is hybrid of Artificial Bee Colony (ABC) algorithm with independent Rough Set Approach. This approach work in two steps:  Identify the subset of attributes independently based on decision attributes.  Identify the final reduct [9]. This approach is used to find out the set of minimal attributes, that is called „reducts‟ to classify objects without degradation in quality of classification and generate minimal length decision rules deep-rooted in a given information system. In order to improve the performance of this method, an element of pruning was introduced. Quick Reduct designate for a dimensionality of n. For the worst-case dataset, evaluations of the dependency function may be performed (n2 +n)/2 times. The dependency of each attribute is estimated and it selects the best candidate [9]. This hybrid algorithm is used to solve both constrained and unconstrained type of optimization problems. This hybrid algorithm uses the rough set theory in a number of real world domains. The performance of this approach is analyzed with five different medical datasets namely Dermatology, Cleveland Heart, HIV, Lung Cancer and Wisconsin and compared with six other reduct algorithms [9].The algorithm given by [9] is as follow: ALGORITHM 5: IQRBEE(Fc,Fd)  Fc, the set of all conditional features;  Fd, the set of decision features.  Bundle the domain  Identify the reduct for each class  Build-up the core reduct and reduct sets  Choose the initial parameter values for ABC  Initialize the population (xi)  Compute the objective and fitness value  Find the optimum feature subset as global  do o Produce new feature subset(vi) o Apply the greedy selection between xi and vi o Calculate the fitness and probability values o Generate the solutions for onlooker o Employ the greedy selection for onlookers o Decide the rejected solution and scouts o Compute the cycle best feature subset o Remember the best optimum feature subset  Repeat up to max_number_of_cycles The suggested algorithm can be used for feature reduction. Let the bees choose the feature subsets arbitrarily and calculate their fitness and realize the simplest one in every iteration. This procedure is continual for variety of iterations to search out the optimum set. As mentioned earlier, when selecting the core reduct, with the remaining attributes at every cluster, the utilized bee produces the feature subset in random. Consider a domain that contains variety of distinctive decision values, then an equivalent range of bees has been chosen because of the population size. From this population half the bees are thought- about as utilized bees and therefore the remaining half is thought-about as onlooker bees. For every used bee, a random subset from one reduct set is allotted. The random sets allotted to all or any the bees are combined to make the feature subset [9]. For instance, contemplate a database that contains ten numbers of conditional attributes (c1,c2,…,c10) and three numbers of call attributes with five hundred records. At the start the records are bundled into three groups supported the choice attribute then the reduct is applied for every cluster. As an example, if we have a tendency to getting the reduct as: R1 = {c1,c3,c4,c8} R2 = {c3,c4,c9} R3 = {c3,c4,c6,c7,c10} From these reducts, the common attributes are chosen as base reduct; during this example, Rc = {c3,c4} and also the remaining attributes are off from every reduct: R1 = {c1, C8}; R2 = { c9}; R3 = {c6,c7,c10} Now three bees are employed to generate a reduct, by choosing random subsets from these reducts and bundling them with the base to find the optimum one. Like:  Rc + Bee_1 = {c8} + Bee_2 = {c9} + Bee_3 = {c7,c10}  {c3,c4,c8,c9,c7,c10} This new reduct is computed using the ABC algorithm. In the second step of the algorithm, for every employed bee, That are exactly half of the number of food sources, a new source is generated as follow: Vi,j = Xi,j +Ø(Xi,j – Xk,j) (6) Where, Øij may be a uniformly distributed real random number among the range [-1,1], k is the index of the answer chosen indiscriminately from the colony (k = int (rand * N) + 1), j = 1,
  • 7. International Journal of Information, Communication and Computing Technology (IJICCT) Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 26 . . .,D and D is the dimension of the problem. When generating vi, this new answer is compared to the existing solution Xi and therefore the utilized bee exploits the higher supply. In the last step of the algorithmic rule, a witness bee chooses a food supply with the likelihood and produces a brand new supply in designated food supply domain. As for utilized bee, the higher supply is set to be exploited. The projected hybrid methodology comes out with a far better optimal reduct than the others; that shows its superior performance [9]. 9. PARAMETER In this paper, the following parameters are used for comparison between hybridization of artificial bee colony algorithms.  Hybrid  Aim  Techniques  Methodology  Dimension  Developer  Linear/Non linear  Constrained/Unconstrained  Known application area The comparative analysis of the different hybrids of ABC optimization technique with Genetic algorithm and Independent rough set is given in form of table-1 at the end of paper. 10. CONCLUSION There are a large number of hybrids of NIC methods to solve complex optimization problems. These hybrids can be classified as per their applicability to the problem and nature of problem. It may be possible that one problem can be solved by many techniques. We can apply more than one optimization techniques to solve the same problem. So in this situation, the complexity plays an important role. It also matter through which technique the best result is obtained whether that result is optimum or not. Now, the different techniques have different complexity. Any problem must be solved with minimum complexity but also provide the optimal solution. Here both ABC-SPSO and HDABC techniques are able to solve constrained and unconstrained problem, ABC-SPSO can be applicable in both 2D/3D where as HDABC is able to solve the problem of D- dimensional space. ABC-SPSO and HDABC both can solve non-linear problems. Different problems can find optimal solution with different techniques. Hybrid Artificial Bee Colony algorithm improve the canonical ABC in solving complex optimization problem whereas Hybrid of Artificial Bee Colony Algorithm with Independent Rough Set Approach for Dimensionality Reduction can solve constrained/ unconstrained optimization problem and use mathematical tools for both feature selection and knowledge discovery. This paper these techniques are presented in an abstract way i.e. in the form of table so that user can choose the technique according to the problem. REFERENCES [1] L. A. Zadeh, “The roles of soft computing and fuzzy logic in the conception, design and deployment of intelligent system,” in Proceedings of the IEEE Asia Pacific Conference on Circuits and Systems, Seoul, Korea, November 1996, pp. 3-4. [2] M. Subotic, M. Tuba, N. Stanarevic, “Parallelization of the Artificial Bee Colony (ABC) Algorithm” Recent Advances in Neural Networks, Fuzzy Systems & Evolutionary Computing, ISSN: 1790-5109,2010, pp. 191-196. [3] K. Price, “An Introduction to differential evolution,” New Ideas in Optimization. D. Corne, M. Dorigo, and F. Glover (Eds.), London, UK: McGraw Hill, 1999. [4] D. Karaboga, “An Idea based on Bee Swarm for Numerical Optimization”, Technical Report, TR-06, Erciyes University Engineering Faculty, Computer Engineering Department (2005). [5] D. Karaboga, A. Basturk, “A powerful and efficient algorithm for numerical function optimization: artificial bee colony (ABC) algorithm”, J. of Global Optimization, Volume 39, No. 0925-5001, 2007, pp. 459-471. [6] A. Banharnsakun, T. Achalakul, B. Sirinaovakul, “The best-so-far selection in Artificial Bee Colony algorithm”, Applied Soft Computing 11 (2011), pp.2888–2901. [7] L. N. de Castro, “Fundamentals of natural computing: An overview,” Physics of Life Reviews, vol. 4, no. 1, pp. 1-36, 2007. [8] X. Wang, X. Z. Gao, and S. J. Ovaska, “Hybrid Nature- Inspired Computation Methods for Optimization” IEEE Soft Computing Research, e-newslatter, December 2009, issue#29 [9] A. Abraham, R. K. Jatoth, A. Rajasekhar, “Hybrid Differential Artificial Bee Colony Algorithm” Journal of computational and theoretical Nanoscience Vol.9,1- 9,2012. [10]K. Price and R. Storn, Differential Evolution – a Simple and Efficient Adaptive Scheme for Global Optimization over Continuous Spaces, Technical Report, International Computer Science Institute, Berkley (1995). [11]Md. El-Abd, “A Hybrid ABC-SPSO Algorithm for Continuous Function Optimization” in IEEE Symposium on Swarm Intelligent System (SIS), 2011. [12]J. Kennedy and R. C. Eberhart, “Particle swarm optimization,” in Proceedings of IEEE International Conference on Neural Networks, vol. 4. Piscataway, NJ: IEEE Press, 1995, pp. 1942–1948. [13]D. Karaboga, B. Akay, “A comparative study of Artificial Bee Colony algorithm”. Applied Mathematics and Computation 214 (2009), pp. 108–132 [14]J. J. Holland, Adaptation in Natural and Artificial Systems, University of Michigan Press, 1975.
  • 8. Comparative study of Hybrids of Artificial Bee Colony Algorithm Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 27 Table 1. Comparative Analysis of Hybrid of ABC Algorithm SN Parameter HABC ABC – PSO HDABC BEEIQR 1 Hybrid ABC and Genetic Algorithm ABC and SPSO ABC and DE ABC and RSAR 2 Aim Improve the canonical ABC in solving complex optimization problem. Aimed at mixing components from both ABC and SPSO in order to have an algorithm that easily solve separable problem as ABC while having a rotationally invariant behavior as SPSO at the same time. Optimization process is applied to each and every individual in the population followed by further improvement using Differential Evolution search Solve constrained/ unconstrained optimization problem and use mathematical tools for both feature selection and knowledge discovery. 3 Technique Single point crossover operator of Genetic Algorithm Population based search and stochastic optimization method. Mutation, Crossover and Selection operator with Differential Evolution. Rough set theory in a number of real world domains. 4 Methodolog y  Set i=0  Initialize the food source positions.  Evaluate the nectar amount (fitness).  While the termination condition are not met  Execute employed bees phase  Calculate the probability P for each food source.  Execute onlooker bees phase  Execute Crossover phase  Execute scout  Initialize and evaluate the swarm  Calculate Max iterations  For each iteration  For every particle i  Update vi and xi  If f(pbesti)< f(xi)  Pbesti = xi  Update gbest  For every particle i  Choose diff random particle and random problem variable  Apply ABC  Initialization While the termination criteria is not met  Apply ABC algorithm  Select best n differential vectors from the population based on the fitness calculated to generate initial population for DE.  For each particle s apply Mutation Crossover Selection  Stop  Cluster domain  Find reduct for each class  Select initial parameter for ABC  Initialize population  Calculate objective and fitness value  Find optimum feature subset  Repeat for maximum no of cycles  Produce initial solution and apply greedy selection  Stop condition
  • 9. International Journal of Information, Communication and Computing Technology (IJICCT) Copyright ©IJICCT, Vol I, Issue II (July – Dec2013): ISSN 2347- 7202 28 bees phase  Memorize the best solution achieved so far.  i=i+1.  Stop condition update rule to pbesti  Update pbesti and gbest  Stop condition condition 5 Dimension 2D/3D 2D/3D D- Dimension space - 6 Developer(s )/ Proposer(s) Xiaohui Yan, Yunlong Zhu, Wenping Zou Mohammed El-Abd Ajith Abraham, Ravi Kumar, A. Rajasekhar Nambiraj Suguna, Keppana Gowder 7 Year 2011 2011 2012 2011 8 Linear/ Non Linear Linear Non – linear Non - linear - 9 Constrained / Unconstrain ed Both Both Both Both 10 Known application area Solve numerical optimization function Function optimization, ANN, Fuzzy control system Real world application: Dynamic load dispatch, harmonic estimation radar tracking Real world application: Medical field 11 References [3] [4] [5] [6] [7] [8] [9] [10] [11]