Application of Genetic Algorithm and Particle Swarm Optimization in Software Testing

IOSR Journal of Computer Engineering (IOSR-JCE)
e-ISSN: 2278-0661,p-ISSN: 2278-8727, Volume 17, Issue 1, Ver. II (Jan – Feb. 2015), PP 75-78
www.iosrjournals.org
DOI: 10.9790/0661-17127578 www.iosrjournals.org 75 | Page
Application of Genetic Algorithm and Particle Swarm
Optimization in Software Testing
Deepti Arora1
, Anurag Singh Baghel2
1,2,
(Gautam Buddha University,Uttar Pradesh,India)
Abstract: This paper will describe a method for optimizing software testing by finding the most error prone
paths in the program. This can be achieved by a meta-heuristic technique, that is by using genetic algorithm
and particle swarm optimization. As exhaustive software testing is not possible where software is tested with all
the possible inputs , those parts of software are also tested which are not error prone..Goal is to generate test
cases using both the algorithms and then comparative study is done between Particle swarm optimization and
genetic algorithm.
Keywords: Software Testing, Meta-heuristics , Genetic Algorithms, Particle Swarm Optimization, Software test
cases.
I. Introduction
Testing is the one of the crucial phase that is performed during software development. It is a primary
technique which is used to gain consumer confidence in the software. It is conducted by executing the program
developed with test inputs and comparing the observed output with the expected one .[1]Software testing is done
to identify any errors or missing requirements in contrary to the actual requirements. It is done to validate the
quality of the software testing using minimal cost and efforts and to generate high quality test cases. It uncovers
the error in the software, including errors in requirement analysis, design document, coding, system resources
and system environments, hardware problems and their interfaces to the software .[2] It is very time consuming
and laborious activity. It is costly and consumes about 50% of the cost of the software development.[3]. Main
goal of testing to find as many faults as possible in a software. As mentioned earlier it a time consuming activity
,as it is impossible to test the software unit with all the combinations of inputs so there is a need to design
minimum no. of test cases that will discover as many faults as possible . However, software testing automation
is not a straight forward process. For years a lot of research have been done in this area to generate test cases
,researchers developed methods like random test case generator, symbolic test data generators and dynamic test
data generators. This paper applied the EST (Evolutionary software testing ) techniquie.EST is the automatic
software test case generation technique which in turn will reduce the cost of development of software. EST
interprets the task of test case generation as an optimization problem and tries to solve it using a search
technique .EST uses an meta-heuristic search technique that is genetic algorithm to convert the task of test case
generation into an optimal problem. Evolutionary Testing search for optimal test parameter combinations that
satisfy a test criterion. Test criterion is represented fitness function that measures how well optimization
parameters that are automatically generated are satisfying the given test criterion .There are variety of
techniques for test case generation but in this paper we will focus on genetic algorithm and particle swarm
optimization technique.
This paper will present the result of our research in into the application of GA and PSO search
approach, to find optimal solution in the software construct.
II. Genetic Algorithm
Genetic Algorithm (GA) are heuristic search algorithm.[4]They are based on the evolutionary idea of
natural selection and genetics. Genetic algorithm is inspired by the Dalton’s theory about evolution that is
survival of the fittest. Genetic algorithm is a part of evolutionary computing which is a growing field of
Artificial Intelligence. GA exploits the historical information to direct the search in region of better performance
with in the search space. [5] In this way competition among individuals for better resources results in fittest
individual dominating over the weaker ones. Genetic Algorithms are more robust hence they are better than
conventional AI(Artificial Intelligence). GA does not break easily even if there is a slightly change in the input
and it also does not get affected by noise. GA offer benefit over typical search of optimization problem in
searching large state space, multimodal search space and n dimensional surface.
Distinct element in GA is individual and population. Individual represents single solution while
population represents set of Solutions which is currently involved in this search process. Each individual is
represented as chromosome. That is an Initial pool of chromosomes. Population size can be few dozens to
thousands. To do optimization fitness function is required to select the best solution from the population and

Application of Genetic Algorithm and Particle Swarm Optimization in Software Testing
discard those not so good solutions. GA consists of following operators that is Selection Operator chromosomes
are selected for cross-over based on the value of the fitness function Cross- Over Operator combine two
chromosomes to produce new chromosomes. Mutation Operator in this value is randomly changed to create new
genes in the individual.
Pseudo code of GA
Initialize (population)
Evaluate Fitness of the population
While (termination condition is met) do
{
Selection(Population)
Cross-over(Population)
Mutation(population)
Evaluate (Population)
}
GA [6] starts with randomly generated population of individuals or chromosomes .Fitness of individual
is calculated based on some fitness function. After the fitness is calculated selection of individuals is done based
on the Roulette –Wheel Selection method i.e an individual having higher fitness value has the more chance of
getting selected .Then Cross over operator is applied to produce new offspring in the population that may have
better characteristics than their parents. Mutation is done to introduce new individual in the population .its is
done by flipping the bit of the chromosomes.
III. Particle Swarm Optimization
Particle Swarm Optimization (PSO) [7] is a relatively recent heuristic search method. It is similar to
GA in the sense that both are evolutionary algorithms .It is one of the meta-heuristics approach that optimizes a
problem and try to improve candidate solution iteratively. PSO is generally used to solve those problems
whose solution can be represented as a point in an n-dimensional space. In PSO potential solution is called
particle. A number of particles are randomly set into motion through this space. Each particle posses its current
position, current velocity ,and its pbest position .Pbest is the personal best position explored so far. It also
incorporates Gbest that global best position achieved by all its individuals. It is a simple approach and it is
effective across a variety of problem domains.
Pseudo code for PSO
The PSO algorithm consists of just few steps, which are repeated until some stopping condition is
met. The steps are as follow:
 Initialize the population of individuals with current position and, velocity.
 Evaluate the fitness of the individual particle (Pbest).
 Keep track of the individual highest fitness (Gbest).
 Modify velocity based on Pbest and Gbest location.
 Update the particle position.
PSO starts with initialization of particle velocity and current position. Here particle is in 2-D space .
Fitness value of the particle is calculated according to function. If the fitness of the particle is better than its
previous value update particle x and y position that is its personal best position. Also if the value is better than
gbest position update global best position of the particle. Apply equations to update the x and y velocity vector
of the particles. Process repeats until termination criteria is met or the optimal solution is found.
IV. Implementation
Genetic Algorithm is implemented in c language. result is as follows. GA approach to problem to
maximize the function (1) F(x)=x3
-x2
. X represent 5 digit unsigned binary integer .i.e x can take value [0,31].
Four randomly generated solution to problem are 14,18,24,30. Fitness of the individual is calculated by the
function F(x).String no. 4 has the maximum value having more chances of getting selected. here N is 4 . One
point Crossover is performed in a couple. New offspring is generated. Mutation is performed. We get new
generation Again the process is repeated till the maximum possible value of the solution is obtained .Luckily in
this case in the second iteration we will get desired value. Whereas In most of the cases GA traps in local
optima. But here We have got the global optimum solution.

Table 1
String
N o.
Initial Population Value of x F(x)=X3
-x2
pi Expected
Count(N*pi)
1 01111 15 3150 0.072 0.289
2 10010 18 5508 0.126 0.505
3 10101 21 8820 0.202 0.809
4 11110 30 26100 0.598 2.395
Sum 43578 1 4
Average 10894 0.25 1
Max 26100 0.598 2.395
Table 2
Before
Crossover
After
Crossover
Mutation
Of element 3 at
position 2
New
Generation is as follows
(Value of x)
01111 01110 01110 14
10010 10011 10011 19
10101 10100 10110 22
11110 11111 11111 31
Particle Swarm Optimization
For Mathematical function (2) val=(x-15)4
-(y-20)2
+9
Particle is initialized with current position and velocity. Program will search for optimal solution
through the movement of particle. Program will keep track of personal best and global best position of the
particle and update its position and velocity .Optimal solution here is to minimize the mathematical function
.PSO is implemented in C language. How the particle moves in the search space to find global best solution is
shown in the Fig 1 & Fig 2. Finally the optimum solution is found when x=15 and y=20.For more complex
problems where it is not possible to easily solve the problem mathematically ,these algorithms can be used.
Figure 1 Movement of particles to find optimum solution
Figure 2 Global best solution

V. Comparison Of GA And PSO
Genetic Algorithm and Particle Swarm Optimization technique are both evolutionary search algorithm
and have been applied in variety of problem domains.[8 &9] GA can be applied to number of problems like N-
queens problem, Travel salesman problem to find optimal solution. GA suffers from the drawback that it traps in
local optima, that is it does not know how to sacrifice short term fitness for long term fitness. It does not keep in
account the global best position of the individual. This drawback is overcome by Particle Swarm optimization
technique which tracks the particle personal best position as well as global best position ,hence it moves toward
global optima without getting trapped in local optima. GA has been popular because of its parallel nature of
search ,and essentially because it can solve non linear ,multi model problems. It can handle both Continuous and
discrete variables where as PSO can easily handle Continuous variables. In comparison to GA ,PSO has few
parameters and it is easy to implement .Calculation of PSO Algorithm is very simple though it is difficult to
implement with problem of non coordinate system .It is problem dependent, Problem having continuous
variables PSO is superior to GA .Many researchers have compared both the technologies .PSO outperforms
GA, and is more effective in general however superiority of PSO is problem dependent.
VI. Conclusion
PSO is relatively recent heuristic approach, It is similar to Genetic algorithm in a way that they both are
population based evolutionary algorithms. Paper is reported on the application of Genetic Algorithm and
Particle Swarm Optimization. Paper described the basic concepts of GA and PSO ,how the test cases are
generated using genetic algorithm and how they are useful in finding the optimal solution to the problem.
Comparative study is done between both the algorithm where GA can be useful and how PSO overcome the
drawback of GA. We also intend to extend the approach described here so that it is useful to solve many
problems like knapsack problem.
References
[1]. B. Beizer,Software testing techniques, Second Edition, Van Nostrand Reinhold, NewYork, 1990.
[2]. Nirmal Kumar Gupta and Dr. Mukesh Kumar Rohil,, Using Genetic Algorithm for Unit Testing Of Object Oriented Software,
IJSSST, Vol.10, No.3.
[3]. S. Desikan and G. Ramesh, ,Software testing principles & practices, Pearson Education, 2007.
[4]. K. F. Man, ,K. S. Tang, and S. Kwong,,Genetic Algorithms: Concepts and Applications, IEEE Transaction on Industrial Electronics
,VOL. 43, NO. 5, October 1996..
[5]. Abhishek Singhal, Swati Chandna and Abhay Bansal, Optimization of Test Cases Using Genetic Algorithm, International Journal of
Emerging Technology and Advanced Engineering ISSN 2250-2459, Volume 2, Issue 3, March 2012.
[6]. Sanjay Singla, Dharminder Kumar, H M Rai3 and Priti Singla, Hybrid PSO Approach to Automate Test Data Generation for Data
Flow Coverage with Dominance Concepts, International Journal of Advanced Science and Technology Vol. 37, December, 2011
[7]. K. Agrawal and G. Srivastava, Towards software test data generation using discrete quantum particle swarm optimization, ISEC,
Mysore,India, pp. 65-68, February 2010
[8]. Andreas Windisch, Stefan Wappler and Joachim Wegener, Applying Particle Swarm Optimization to Software Testing,
GECCO’07, July 7–11, 2007, London, England, United Kingdom Copyright 2007 ACM 978-1-59593-697-4/07/0007.
[9]. Praveen Ranjan Srivastava1 and Tai-hoon Kim2, Application of Genetic Algorithm in Software Testing,, International Journal of
Software Engineering and Its Applications Vol. 3, No.4, October 2009.

Application of Genetic Algorithm and Particle Swarm Optimization in Software Testing

More Related Content

What's hot (16)

Similar to Application of Genetic Algorithm and Particle Swarm Optimization in Software Testing (20)

More from IOSR Journals (20)

Recently uploaded (20)

Application of Genetic Algorithm and Particle Swarm Optimization in Software Testing