Evaluating The Effectiveness of Regression Test Selection Using Ant Colony Optimization
Evaluating The Effectiveness of Regression Test Selection Using Ant Colony Optimization
International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
Web Site: www.ijettcs.org Email: [email protected]
Volume 4, Issue 3, May-June 2015
ISSN 2278-6856
Evaluating the Effectiveness of Regression
Test selection using Ant Colony Optimization
1.
Anjali Choudhary, Tarun Dalal2
M.Tech Scholar CBS Group of Instituion, Maharishi Dayanand University, Jhajjar, Haryana, India
Asst. Professor CBS Group of Instituion, Maharishi Dayanand University, Jhajjar, Haryana, India
Abstract
Regression testing in an important phase in Software
Development Life Cycle. This activity is generally performed
in Software Maintenance phase. This activity is performed
whenever a major or minor change is made to the application
code on demand of the client or end under. These
modifications can sometimes cause the application to work in
inappropriate way. In order to catch bugs and errors, testers
have many test cases to test the application. But in order to
find out the bugs, all the test cases need to be run. Running
each and every test cases is not possible every time dues to
time and cost constraints. To select and prioritize the test
cases, we are using algorithm, Ant Colony Optimization.
Keywords: Regression test case, selection, Prioritization,
ACO, implementation.
1. INTRODUCTION
Regression testing process is carried out in maintenance
phase of the software and it is one of the important
phases. Regression testing is performed whenever the
application code is modified. To test the modified
changes whether they are working correctly or not,
Regression testing is done. Sometime it is possible that
due to budget and time constraint, it is not possible to rerun all the test cases in the test suite to test the modified
code. In that scenario, we use test case minimization,
selection and prioritization technique for Regression
testing.
Regression testing is performed in time constraint
environment where testing is done for a limited time. In
year 2010, Singh et al [7], also proposed a time constraint
prioritization technique using Ant colony Optimization
(ACO). The results presented in the paper provide
motivation for implementing algorithm. In this paper, we
will use tool Ant Colony Optimization which is generally
used for solving Travelling Sales Person (TSP) problem.
But we will use this tool for Regression test case selection
and prioritization.
In this paper, Time and Space complexity of the
algorithm is measured which provided further motivation
to implement the technique. The analysis represents the
effectiveness of ACO technique for test case selection and
prioritization.
2. ANT COLONY OPTIMIZATION
Ant Colony Optimization technique is a set of search
algorithms of Artificial Intelligence for optimal solutions.
This technique was proposed by Colorni, Dorigo and
Maniezzo [2][3][4]. The iconic parts of ACO are ANTS.
Ants are blind and are small in size, they are still able to
search for their food and reach their home using shortest
path. They make use of their pheromones and antennas to
be in touch with each other and find the optimal path.
ACO is inspired from the behavior of ants.
ACO deals with two important processes, namely
Pheromone deposition and pheromone evaporation.
Pheromone deposition is the phenomena in which as the
ant move from one place to another, it releases a
pheromone liquid. Pheromone evaporation is the
phenomena that over the time, deposited pheromone have
some time period after which it evaporates and vanishes.
This pheromone trail is followed by other ants by
smelling which tends to follow the path with maximum
pheromone trail. Point of interest is that the ant the food
source and come back to the source even when they dont
have eyes. Thus, ant on the shortest path is the first to
come back to its nest. The process of path covering by
ants is shown in Fig 1. This can be illustrated with an
example demonstrated below:
Figure 1. Regression testing process
Volume 4, Issue 3, May June 2015
Page 60
International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
Web Site: www.ijettcs.org Email: [email protected]
Volume 4, Issue 3, May-June 2015
ISSN 2278-6856
Figure 2 ACO example
Suppose there are 3 paths P1, P2 and P3, from the
source(ant nest) to the destination (food source). When
ant move from its nest it can choose any of the path P1,
P2, or P3. After collecting food from the source ants move
back to their nest. Ants on the shortest path will b the one
who will reach first at their nest. Therefore more amount
of pheromone will deposit on this path. New ants starting
from the nest may tend to follow this shortest path and
deposit more pheromone on this path. Thus, it is observed
that most of the ants follow the best path and this best
path has maximum pheromone deposited. This
phenomenon of finding the best path is called Ant Colony
Optimization.
algorithm in test case selection and prioritization w.r.t to
time, we computed the complexity of the ACO. The
Algorithm as shown in Fig.3 has a run time bounded by
the time required to generate artificial ants al to an in
step1, also the number of iteration for three nested loop in
step2. Suppose t be the regression test suite with |t| no.
of total test cases in it. Let tc be the time input by the
user.
Generation of artificial ants in step1 is O(|t|) operation[5].
The inner most loop do-while calls select test case() till
all faults are covered. This can repeat for maximum |t|
times as there are at most |t| test cases in the test suite t.
Thus, in the best case only 1 call to select the test case
needs to be made to select test case(). The second nested
loop i.e. for loop repeats for |t| iteration.
The outermost Do while loop repeats until the execution
time of the selected test case for every iteration increase
beyond the user defined constraints Tc, which is constant
and independent of the size of the test suite. Hence, the
overall complexity of the ACO is O(TC. |t|2) for worst
case, which is equivalent to O(|t|2). Select test case() itself
takes constant time to execute and is independent of the
size of the test suite or time constant input.
Thus, the best case complexity for the ACO algo for test
selection and prioritization comes out to be O(Tc|t|
equivalent to O(|t|), as the function select_test_case() is
called only once to cover all faults in the application or
code. This complexity is acceptable as compared to other
time aware prioritization as NP-complete.
3. ACO ALGORITHM
Figure 3 Flow diagram of Ant Colony Optimization
representing Ants behavior
Through this behavior of ants, Dorigo proposed Ant
Colony Optimization (ACO) in year 2006 [3]. ACO has
been used to solve many problems like Travelling sales
person problem, data mining, vehicle routing, test data
generation. As proposed by Singh et al, ACO can also be
used for prioritizing the test cases.
2.1Complexity of ACO algorithm
The whole effort of building an algorithm is wasted if the
algorithm itself takes more time to execute all test cases
in regression test suite. To prove the efficiency [1] of the
Volume 4, Issue 3, May June 2015
Figure 4 ACO Algorithm for test case selection and
prioritization proposed by Singh et al
Page 61
International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
Web Site: www.ijettcs.org Email: [email protected]
Volume 4, Issue 3, May-June 2015
ISSN 2278-6856
In ACO pheromones were updated only when all the ants
completed travel for a given iteration. Ant Colony system
is bit different. After ants moves from one test case to
another, the pheromone on the travelled path is reduced
by some value [6].
The algorithm has a parameter
. The
parameter sets the boundary between two rules to select
the next test case. On each step, algorithm generates a
random number
. If
, the next test case
to move is selected according to (1),
; else index J of the next test case to move is:
Where
I is the index of the current test case
J is the index of the test case
is a set of test cases adjacent to the current test case.
il is the amount of the pheromone on the path cil.
algorithm. The selection threshold
reset parameter.
If,
is a
All pheromones on the path between all test cases are
reset to
and search for the new shortest path.
3.2.1Implementation
Tool used in this paper is Ant Colony Algorithm. This
tool is basically used for solving Travelling Sales Person
problem. We have used this tool for test case
prioritization and selection as well. The shortest path
covered will be given the priority among many test cases;
It is obvious that the test case which is covered in the
shortest distance will be executed in minimum time.
Testing is performed to recover the faults in the
application. So it is obvious that every selected test case
would recover some fault or pass the test case.
Ant Colony Algorithm tool used in this research looks as
follow[6]:
, Q is a constant, dil is the distance between
the test case I and l
is the algorithm parameter
After ant moves from one test case to another, pheromone
on the path cil are evaporated and updated according to
After the first ant has made the Kth step and has modified
a pheromone, it does not begin the next step. It waits until
all the ant completes there Kth step and modify the
pheromones on their travelled path. In this process, the
n+1 ant waits for the update by the ant n and so on.
3.1Evaporation and pheromone update
After all ants have completed iteration, pheromone update
and global evaporation are applied to the paths traversed
by the ants. In ACO, only the best to date any is allowed
to update pheromone on the path. Evaporation is allowed
only on path which are visited by the ants. Evaporation
used equation 3,
(3)
And (2) and (4) are used for Global pheromone update.
(4)
The equations will only be applied to paths visited by the
selected ants only.
3.2Pheromone Reset
If ants are struck in loop, there is an option to reset all
pheromones to start values and continue run. This option
might be enabled or disabled for all implemented
Volume 4, Issue 3, May June 2015
Figure 5: Ant Colony Algorithm Tool
We have provided the input as:
Method of test case generation can be selected at
Random, Automatic or Manual. We have selected the
test case generation as Random.
Test case extent is the distance between the test cases
coordinates.
Numbers of test cases are set to 10.
Max test cases are 1000.
Alpha i.e. Power of pheromone is set to 3.
Beta i.e. weight on edges is set to 2.
Evaporation is global evaporation of the pheromone
0.100.
After providing the input values we ran the application.
The result is generated in the form of graph and a
Page 62
International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
Web Site: www.ijettcs.org Email: [email protected]
Volume 4, Issue 3, May-June 2015
ISSN 2278-6856
diagram is created after the run as shown in fig 5.
2, 8
4
9
5
3
0
4.01
4.528
4.668
5.089
5.36
7.281
After the run, test case with shortest distance travelled
will be the best test case and test case which took longest
distance to travel will be the worst test case. Eventually
best test case took the minimum time as it traveled the
shortest path and worst test case took the maximum time
as it travelled the longest path. On the basis of this
analysis test cases selection and prioritization is done.
A graph is also generated at the end of the run which is
shown in graph1. This graph is generated from Random
placement of test cases.
Figure 6 Final covered paths by ants
It has displayed the path which is best for iteration. Ant 5
travelled the best path to iteration. As we are representing
the ants to the test cases, then test case 5 will take
minimum time to execute. This test case 5 will be selected
and will be prioritized. Similarly we have ranked all the
test cases in the decreasing order of the distance travelled.
Epoch is the particular time of period which is set to 1000
by default. The results are displayed in the table1 below:
Table 1 Prioritized order of test cases using Random
Placement
Graph1. Graph representing the best and best to date ant
Prioritized
case
5
1
8
2
0
7
9
3
6
4
Test
Distance
travelled
3.483
3.848
3.87
3.89
4.45
4.459
4.555
4.755
4.792
5.198
We have compared the results using different test
generation methods as Automatic and Manual.
In automatic generation, ants are made to travel in circle.
There are chances that two or more ants may travel the
same distance. In that case it becomes difficult to
prioritize the ants i.e. test case. Similar problem is faced
in Manual selection of test case method generation.
Therefore, preference is made for Random selection.
Table 2 Prioritized order of test cases using Automatic
Placement
Prioritized
case
1,6,7
Test
Distance
travelled
3.129
Volume 4, Issue 3, May June 2015
This graph represents the best and best to date distance
travelled by each artificial generated ant. It has two
values currently representing Best and best to date
distance of ant.
4. CONCLUSION AND FUTURE WORK
In this paper we have presented ACO technique for test
selection and prioritization which leads to optimal
solution for selecting test cases. Efficiency of the
algorithm is also computed which has encouraged the use
of ACO in time constraint test selection and
prioritization. In future we wish to apply the ACO on big
and more complex system problems.
References
[1] Y.Singh, A.Kaur, B.Suri, Test case prioritization
using ant colony optimization, ACM SIGSOFT
Software Engineering Notes, Vol.35 No.4, pages 1-7,
July 2010
[2] A. Colorni, M. Dorigo, and V. Maniezzo,
Distributed optimization by ant colonies,
Proceedings of ECAL'91, European Conference on
Artificial Life, Elsevier Publishing, Amsterdam,
[3] M. Dorigo, Optimization, learning and natural
algorithms, Ph.D. Thesis, Politecnico diMilano,
Milano, 1992.
Page 63
International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
Web Site: www.ijettcs.org Email: [email protected]
Volume 4, Issue 3, May-June 2015
ISSN 2278-6856
[4] M. Dorigo, V. Maniezzo, and A. Colorni, The ant
system: an autocatalytic optimizing process,
Technical Report TR91-016, Politecnico di Milano
(1991).
[5] T.H.Cormen, C.E.Leiserson, R.L.Rivest and C.Stein,
Introduction to Algorithms, PHI Publications, 2009
edition.
[6] www.codeproject.com
[7] Y.Singh, A.Kaur, B.Suri, Test case prioritization
using ant colony optimization, ACM SIGSOFT
Software Engineering Notes, Vol.35 No.4, pages 1-7,
July 2010.
AUTHOR(S)
Anjali Choudhary received the B.tech in Information
technology from Rajasthan Technical University. She is
now pursuing M.tech in Computer Science and
Engineering from CBS group of institution from
Maharishi Dayanand Univeristy, Haryana, India.
Tarun Dalal is Asst Professor in department of Computer
Science and Engineering, CBS group of institution,
affiliated from Maharishi Dayanand University, Haryana,
India.
Volume 4, Issue 3, May June 2015
Page 64
