2019 International Conference on Machine Learning, Big Data, Cloud and Parallel Computing (Com-IT-Con), India, 14th -16th

Feb 2019

Genetic Algorithm- A Literature Review

Annu Lambora, Kunal Gupta, Kriti Chopra
Department of Computer Science and Engineering
ASET, Amity University
Noida, India
[email protected], [email protected], [email protected]

Abstract: Genetic Algorithm (GA) may be attributed as method

for optimizing the search tool for difficult problems based on
genetics selection principle. In additions to Optimization it also
serves the purpose of machine learning and for Research and
development. It is analogous to biology for chromosome
generation with variables such as selection, crossover and
mutation together constituting genetic operations which would
be applicable on a random population initially.GA aims to yield
solutions for the consecutive generations. The extent of success in
individual production is directly in proportion to fitness of
solution which is represented by it, thereby ensuring that quality Fig. 1. Processing of a Data
in successive generations will be better. The process is concluded In optimization we need to give insert in way so that it can
once an GA is most suitable for the issues that need optimization provide the “best” output. The precision of “best” depends
associated with some computable system.. John Holland may be on the different types of the problem [3]. If we consider
regarded as funding father of original genetic algorithm and is mathematical terms, it refers to minimizing or maximizing
attributed to year 1970’s as funding date .Additionally a random more than one object functions, by supplying different input
search method represented by Charles Darwin for a defined guideline. The set of all possible inputs and values provider
search space in order to effetely solve a problem. In this paper, the possible solution to make up the search space. In this
what is genetic algorithm and its basic workflow is discussed how
search space, there are point or more than one point lies
a genetic algorithm work and what are the process is included in
which give the optimal solution. The focus of optimization is
this is also discussed. Further, the features and application of
genetic algorithm are mentioned in the paper. to discover that point or set of points in the search space.

Keywords- Genetic Algorithm, Inheritance, Mutation,

Selection, Crossover II. LITERATURE REVIEW
Genetic algorithm is of great use in order to find out shorted
path by application of path encoding in graph converting into
I. INTRODUCTION
One may consider the biological evolution aspect and Natural chromosomes. The approach was well tested by gen metal
selection theory as basis of defining genetic algorithm, with (1997) for a wide size range of nuclear ( 6 to 70) and from
application for search optimization in computing and artificial varied edges ( 10 to 211).The final result by the application
of genetic algorithm can be used to find optimal result
intelligence Large and unorganized data search is most
instantly and with a very high accuracy.
effective using genetic algorithms. Generally complex issues
which are constrained and unconstrained are effectively
Application of genetic algorithm as a method of search for
solved using genetic algorithm. Genetic algorithm is mostly optimization problem in diverse manner was examined by
used for machine learning and problem optimization purposes Jang Sung Chun (1998). Evolutionary algorithm and genetic
and many others also. Optimization is the technique to make algorithm were compared on varied problems of optimization
something better than other technique .In optimization some and the outcomes resulted the outer performance of genetic
sets of input are provided and based on the input we get the algorithm.
output.

978-1-7281-0211-5/19/$31.00 2019 ©IEEE 380

 III. GENETIC ALGORITHM AND ITS WORKFLOW process for this to be executed is the selection of a crossover point
in random manner, taking range from first parent point till this
Genetic Algorithms (GAs) are generally search based
point . The illustration of cross over point cab be shown as below
algorithms established on the concepts of natural selection
diagram:
and heredity. GA is a sub division of a much bigger area of
computation which is called as Evolutionary computation.
In GAs, for a given problem, we have a varied and Chromosome 1 11011 | 00100110110
numerous solutions. Chromosome 2 11011 | 11000011110
The solutions obtained then go through recombination as Offspring 1 11011 | 11000011110
well as process of mutation (similar to biological genetics), Offspring 2 11011 | 00100110110
lead to generating new offspring’s and the process get
replicates for multiple generations [4]. Each single is given a Fig. 3. Crossover of Chromosome
unique fitness vale which is decided by the value of its There are many different ways to make crossover, like as we can
objective function and the individuals who are more fir have select many other crossover points. Crossover can be more
a greater possibility of mating to produce further Fit complex and intricate. It depends mainly on encoding of
individuals. This process ensures to create Fitter individuals chromosome. For improving the performance of genetic algorithm
or better solutions in succeeding generations, which select crossover for the special problems.
continues till it reach final destination criterion.
a) Mutation
Genetic algorithms to a great extent are probability based
After the crossover is performed, next step is mutation. Mutation is
criterion in nature, on the contrary they work well as against
deliberate to stop falling of all solution in the population into a
local random search ( Which uses random solutions, unable local optimum of the solved problem. Offspring results from
to identify best possible solutions), due to the fact that crossover randomly changes by mutation operation. In binary
historic data gets into it making it complex and sophisticated encoding, we can switch some randomly selected bits from 0 to 1
in nature or 1 to 0.Mutation can be adorn as given below:
Operations perform in GA:

Original offspring 1 1101111000011110

• Selection
• Crossover Original offspring 2 1101100100110110
• Mutation Mutated offspring 1 1100111000011110
Mutated offspring 2 1101101100110110
a) Selection (Encoding of a Chromosome)
Fig.4: Mutation of Offspring
A chromosome will be in a format that hold information The process of mutation (as well as crossover) totally depends on
related to solution that it personify. Widely used way of the encoding of chromosomes.
encoding is a binary string format. After that chromosome
will look like this. b) Crossover and Mutation Probability
Genetic algorithm consist of two basic parameters namely
Chromosome 1 1101100100110110 crossover probability and mutation probability.

Chromosome 2 1101111000011110 c) Crossover probability

Fig. 2. Chromosomes This may be attributed to the extent of frequency of crossover. If
Every one of the chromosomes can be plot by a binary string. offspring’s are duplicate of parents that means there is no
Each of the bit accommodate in the string is also responsible crossover. In case of crossover, the two part of parent’s crossover
containing some aspects or standard of the solution. result into an offspring. A probability of 100% will be there for
offspring from crossover. On the other hand, if whole of new
b) Crossover generation is produced as carbon copy of chromosomes from
Once we are confirmed on the selected coding to be used, we earlier population (doesn’t necessarily indicate that new generation
can move forward to crossover operation. Crossover will be same) then it will be 0%.Basically crossover is done so that
functions on part portion of genes from parent chromosomes chances of new chromosomes consisting of good character of the
which results in creation of a new offspring [5]. The simplest old chromosomes so that new crossover will be batter then old one.
381
It is batter to left some portion of old population in order for x No need of Imitative information ( which don’t exist for
survival till the next generation. most of the real life problems)

x A more comprehensive and optimum in comparison to the
d) Mutation probability
It may be described as the chance of chromosome parts to get primitive methods

mutated. In the absence of mutation, offspring are produced x Possess a well aligned capabilities
with no change immediately preceding crossover. 
x Enhancing both continuous as well as distinct functions
The value of mutation probability is in direct proportion to along with multi-purpose problems
chromosome change. If value is 100%, 
x Aims at delivering Best solutions instead of a single
then there would be whole chromosome change, for 0% value
solution
no chromosome change is there. Mutation process helps in 
preventing GA from the downfall into local extremes [6]. x Having a satisfactory answer for the problems, which
The frequency of occurrence of mutation should be low, improves with time
otherwise GA will transform into a random search 
x Works best in cases of large search space area possessing
multiple parameters into it.

V. APPLICATIONS

A. Network Routing Protocol-

By using optimum coding scheme, route for shortest path can be

found for Length of chromosomes is directly proportional to
quantity of nodes for a network. By using Dijkstra’s algorithm.
Search can be done for optimal and shortest path using MATLAB
environment. The algorithm effectively solved 10 nodes network
as source node for first time. Additionally for tackling same
problem developed GA is used. The results obtained were same as
using Djikstra’s algorithm signifying the potential of this proposed
algorithm. For future applications, more research need to be done
on decreasing chromosome length for developed GA especially for
Fig. 5. Genetic Algorithm Flow Chart cases involving multi node networks [8].

First step aims at certain of the initial population and after B. Image Processing-
that analyzing fitness factor of all the chromosomes by the Image processing is critical area where genetic algorithm proved to
application of fitness function. be a boon and it is deemed to the
Seceding first step, chromosomes possessing fitness values most successful technique for large space Multiple variables
similar to threshold will be selected as next generation ranging from image contrast, enhancing detail to advanced filters
parent.noe check this criteria, if it lies within the specified and deformable parameters are effectively solved by application of
criteria then we need to stop, else any of two strings can be this paradigm. A comprehensive and vigorous search can be
selected from initial population and after that crossover can performed using this algorithm eradicating trap in local extremes
be performed on it leading to offspring formation and again and also to derive optimum global search results. Optimization
checking the criteria of star function, if yes then we need to results are dependent upon chromosome encoding patterns,
stop, else the process need to be continued until the problem involvement of fitness function as well as genetic operators.
is solved satisfactorily. However, image segmentation quality can be enhanced by
parameter selection in an optimized manner. Image enhancement
IV. ADVANATGES denotes better digital image quality, which is not concerned with
A number of advantages made GAs to be applicable in source of degradation
multiple areas -
C. Data Mining-

382
Inventory management is one of the critical areas in supply 1. GA output may be considered as an approach on probability
chain management. Several methods in inventory and there is a not standard set rule governing GA .This shows
management can be used supply chain cost by effectively that solution is not guaranteed as per our requirement. Instead,
controlling the inventory, thereby ensuring that members of we need to consider that whatever solutions we get are best
supply chain won’t be affected by discrepancies in stock possible which are delivered by GA
levels which may be short or excess. One of the major issues
is drastic change in stock level while considering the 2. Multiple variable parameters work with random numbers in GA
implementation. This paper aims at finding optimum .For example probabilities of crossover and different types of
approach in order to minimize supply chain cost mutation, probability of selection and method of selection and
number of iterations.
D. Computer Games-
Genetic algorithm finds its application for computer games. It 3. An advantage as well as disadvantage of GA is that it aims at
creates a realistic experience for players, main aim for player solving through evolution. Evolution is not in uni-direction, for
is to discover sequence of steps which if repeated in different natural life it evolves away from bad situation and not towards
games, will create a winning situation for players, there a good solution. It can lead to evolution of species into
won’t be any challenge. On the contrary, by using genetic evolutionary dead end. GA can provide sub-optimal solution
algorithm technique effectively player can avoid repetition of and we not be aware of it.. Convergence principle is a big
past mistakes, the game would be more popular among other phenomenon in GA For example in order to start GA search by
players. giving an input by user for getting results, certain values are
obtained which are the optimal solution at that particular point
Genetic algorithms require the following components: of time. This too proves to be a major setback. But the result
obtained may not be the optimal one and better results could
x The way for displaying challenges expressed in the form be possible in real life situation
of solutions ( Like Synchronizing soldiers for an attack
in a strategy based game) The cause is that, sometimes in a search process of GA there may

x A function based on evaluation or fitness to know the be a local optimal point as descent in the local optimal point it can’t
quality of an occurrence ( example computation of go further to search for global optimal. Which is actually required
destruction done to opposition in such situation of by us. This property of descent within the local optimal space is
attack) known as convergence. This is a pitfall in GA searching procedure

In order to measure the quality for an entity the fitness VI. CONCLUSION
function incorporates a mutated function, which in-turn can
be customized to the issue domain. In many instances Theoretical concepts of Genetic Algorithms are discussed in detail
specially containing code optimization, the fitness function in the introduction part initially. It is analogous to biology for
can be attributed as system timing function. After defining chromosome generation with variables such as selection, crossover
Fitness function as well as genetic representation, a genetic and mutation together constituting genetic operations which would
algorithm will incorporate initial members which were be applicable on a random population initially. The genetic
discussed previously, and after that improving with the algorithm may be described as a population oriented search and
mutation application in a repetitive manner, crossover, optimization approach that resembles the natural evolution
reversal and selection operators ( synonymous to the problem process. Genetic algorithm functioning is closely effected by two
domain) main natural evolution concepts – First in natural selection and
Practical obstacles and challenges of Genetic Algorithm: other is genetic dynamics involving different genetic operations
Despite of having multiple useful hypothetical cases found to like crossover, mutation etc. Genetic Algorithms are highly
be efficient for optimum solutions for cell formations applicable in immeasurable applications like Search, Optimization,
problems, the practical aspects instead of theoretical Decision Making, Machine Learning, Robotics and many more .In
approach , genetic algorithm use is still not clear [9]. this paper we discussed the basic work flow chart of genetic
Feedback from industrial engineers and researchers shows a algorithm and all the function of it in detail after that some
big gap. Below are some important points to realize practical advantages of using genetic algorithm are discussed and some
blockades and challenges encountered by industrial engineers genetic algorithm application are discussed and in last the issues
and researchers for bridging this gap and challenges are discussed.

383
REFERENCES
[1] A. Verma and Archana, ―A Survey on Image Contrast
Enhancement Using Genetic Algorithm,ǁ International
Journal of Scientific and Research Publications, Volume
2, Issue 7, 1 ISSN 2250-3153, pp.2 ,July 2012.
[2] H. D. Mills, M. Dyer, and R. C. Linger, "Cleanroom
software engineering," IEEE Software, vol. 4, p. 19, 1987
[3] Ting, Chuan-Kang (2005). "On the Mean Convergence Time of
Multi-parent Genetic Algorithms without Selection". Advances in
Artificial Life: 403–412. ISBN 978-3-540-28848-0.
[4] E. Dum bill, “Making sense of big data,” Big Data, vol. 1, no.
1,pp. 1–2, 2013.
[5] E. W. Dijkstra, "A Note on Two Problems in Connexion with
Graphs", Numerische Mathematlk l, 269 – 271, l 959.
[6] N. Selvanathan and W. J. Tee, "A Genetic Algorithm Solution to
Solve The Shortest Path Problem OSPF and MPLS", Malaysian
Journal of Computer Science, Vol. 16 No.1, pp. 58-67, 2003.
[7] Goldberg, David E. (1989). Genetic Algorithms in Search
Optimization and Machine Learning. Addison Wesley. P.41
ISBN0-201-15767-5.
[8] B. Fan and J. Luo, “Spatially enabled emergency event analysis
using a multi-level association rule mining method,” Natural
Hazards, vol. 67, no. 2, pp. 239–260, 2013.
[9] F. C. Su and W. L. Wu, “design and testing of a genetic
algorithm neural network in the assessment of gait patterns”,
institute of biomedical engg. National cheng Kung University, pp
67-74, February 2000.
[10] A. Verma and Archana, ―A Survey on Image Contrast
Enhancement Using Genetic Algorithm,ǁInternational Journal of
Scientific and Research Publications, Volume 2, Issue 7, 1 ISSN
2250-3153, pp.2July 2012.
[11] Gervautz, Traxler: Representation and Realistic Rendering of
Natural Scenes with Cyclic CSG graphs, accepted for publication
in Visual Computer, 1995
[12] Mardle, S. and Pascoe, S. 1997. A review of applications to
fisheries using multiobjective programming techniques.
CEMARE Research Paper, PI 17, University of Portsmouth, UK,
24pp.
[13] Jaffry S., Pascoe S. and Robinson C. (1997); 'Long run price
flexibilities for high valued species in the UK: a cointegration
systems approach', CEMARE Research Paper, CEMARE,
University of Portsmouth, UK, Pill, 12pp.
[14] WorldBank, Manufacturing, Value Added (% of
GDP),2016,http://data.worldbank.org/indicator/NV.IND.MANF.
ZS/countries?display=graph.
[15] Grobler, A. P. Engelbrecht, S. Kok, and S.
Yadavalli,“Metaheuristics for the multi-objective FJSP with
sequencedependent set-up times, auxiliary resources andmachine
downtime, ” Annals of Operations Research, vol. 180, no. 1, pp.
165–196,2010.

384